{"thread":{"id":"49277","subject":"What's cooking in git.git (Sep 2018, #01; Tue, 4)","startedAt":"2018-09-04T22:36:18Z","lastAt":"2018-11-25T14:14:52Z","messageCount":19,"participants":["Junio C Hamano","Ævar Arnfjörð Bjarmason","Eric Sunshine","Johannes Schindelin","Duy Nguyen","SZEDER Gábor","Jeff King","Torsten Bögershausen"],"isPatch":false,"patchVersion":null,"patchTotal":null},"messages":[{"id":"357389","messageId":"xmqqmusw6gbo.fsf@gitster-ct.c.googlers.com","threadId":"49277","inReplyTo":null,"subject":"What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2018-09-04T22:36:11Z","receivedAt":"2018-09-04T22:36:18Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Here are the topics that have been cooking. Commits prefixed with\n'-' are only in 'pu' (proposed updates) while commits prefixed with\n'+' are in 'next'. The ones marked with '.' do not appear in any of\nthe integration branches, but I am still holding onto them.\n\nGit 2.19-rc2 is out. Hopefully the tip of 'master' is more or less\nidentical to the final one without needing much updates.\n\nYou can find the changes described here in the integration branches\nof the repositories listed at\n\n http://git-blame.blogspot.com/p/git-public-repositories.html\n\n--------------------------------------------------\n[Graduated to \"master\"]\n\n* ab/portable (2018-08-27) 6 commits\n (merged to 'next' on 2018-08-27 at 37640e66ef)\n + tests: fix and add lint for non-portable grep --file\n + tests: fix version-specific portability issue in Perl JSON\n + tests: use shorter labels in chainlint.sed for AIX sed\n + tests: fix comment syntax in chainlint.sed for AIX sed\n + tests: fix and add lint for non-portable seq\n + tests: fix and add lint for non-portable head -c N\n (this branch is used by ab/portable-more.)\n\n Portability fix.\n\n\n* ab/portable-more (2018-08-29) 2 commits\n (merged to 'next' on 2018-08-31 at d7b44993e4)\n + tests: fix non-portable iconv invocation\n + tests: fix non-portable \"${var:-\"str\"}\" construct\n (this branch uses ab/portable.)\n\n Portability fix.\n\n\n* ds/commit-graph-lockfile-fix (2018-08-30) 1 commit\n (merged to 'next' on 2018-09-04 at 0876e6ddcc)\n + commit: don't use generation numbers if not needed\n\n \"git merge-base\" in 2.19-rc1 has performance regression when the\n (experimental) commit-graph feature is in use, which has been\n mitigated.\n\n\n* en/directory-renames-nothanks (2018-08-30) 3 commits\n (merged to 'next' on 2018-08-31 at 91d663d688)\n + am: avoid directory rename detection when calling recursive merge machinery\n + merge-recursive: add ability to turn off directory rename detection\n + t3401: add another directory rename testcase for rebase and am\n\n Recent addition of \"directory rename\" heuristics to the\n merge-recursive backend makes the command susceptible to false\n positives and false negatives. In the context of \"git am -3\",\n which does not know about surrounding unmodified paths and thus\n cannot inform the merge machinery about the full trees involved,\n this risk is particularly severe. As such, the heuristic is\n disabled for \"git am -3\" to keep the machinery \"more stupid but\n predictable\".\n\n\n* es/chain-lint-more (2018-08-29) 1 commit\n (merged to 'next' on 2018-08-31 at d456090b62)\n + chainlint: match \"quoted\" here-doc tags\n\n The test linter code has learned that the end of here-doc mark\n \"EOF\" can be quoted in a double-quote pair, not just in a\n single-quote pair.\n\n\n* es/freebsd-iconv-portability (2018-08-31) 1 commit\n (merged to 'next' on 2018-09-04 at 52baa37dd7)\n + config.mak.uname: resolve FreeBSD iconv-related compilation warning\n\n Build fix.\n\n\n* pw/rebase-i-author-script-fix (2018-08-07) 2 commits\n (merged to 'next' on 2018-08-31 at 7b9f485407)\n + sequencer: fix quoting in write_author_script\n + sequencer: handle errors from read_author_ident()\n\n Recent \"git rebase -i\" update started to write bogusly formatted\n author-script, with a matching broken reading code. These are\n fixed.\n\n--------------------------------------------------\n[New Topics]\n\n* ab/fetch-tags-noclobber (2018-08-31) 9 commits\n - fetch: stop clobbering existing tags without --force\n - fetch: document local ref updates with/without --force\n - push doc: correct lies about how push refspecs work\n - push doc: move mention of \"tag \" later in the prose\n - push doc: remove confusing mention of remote merger\n - fetch tests: add a test for clobbering tag behavior\n - push tests: use spaces in interpolated string\n - push tests: make use of unused $1 in test description\n - fetch: change \"branch\" to \"reference\" in --force -h output\n\n The rules used by \"git push\" and \"git fetch\" to determine if a ref\n can or cannot be updated were inconsistent; specifically, fetching\n to update existing tags were allowed even though tags are supposed\n to be unmoving anchoring points. \"git fetch\" was taught to forbid\n updates to existing tags without the \"--force\" option.\n\n Will merge to and cook in 'next'.\n This is a backward incompatible change but in a good way; it may\n still need to be treated carefully.\n\n\n* es/worktree-forced-ops-fix (2018-08-30) 9 commits\n - worktree: delete .git/worktrees if empty after 'remove'\n - worktree: teach 'remove' to override lock when --force given twice\n - worktree: teach 'move' to override lock when --force given twice\n - worktree: teach 'add' to respect --force for registered but missing path\n - worktree: disallow adding same path multiple times\n - worktree: prepare for more checks of whether path can become worktree\n - worktree: generalize delete_git_dir() to reduce code duplication\n - worktree: move delete_git_dir() earlier in file for upcoming new callers\n - worktree: don't die() in library function find_worktree()\n\n Various subcommands of \"git worktree\" take '--force' but did not\n behave sensibly, which has been corrected.\n\n\n* jk/patch-corrupted-delta-fix (2018-08-30) 6 commits\n - t5303: use printf to generate delta bases\n - patch-delta: handle truncated copy parameters\n - patch-delta: consistently report corruption\n - patch-delta: fix oob read\n - t5303: test some corrupt deltas\n - test-delta: read input into a heap buffer\n\n Malformed or crafted data in packstream can make our code attempt\n to read or write past the allocated buffer and abort, instead of\n reporting an error, which has been fixed.\n\n Will merge to 'next'.\n\n\n* jc/rebase-in-c-9-fixes (2018-09-04) 1 commit\n - rebase: re-add forgotten -k that stands for --keep-empty\n (this branch uses ag/rebase-i-in-c, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final.)\n\n\n* jk/pack-objects-with-bitmap-fix (2018-09-04) 4 commits\n - pack-bitmap: drop \"loaded\" flag\n - traverse_bitmap_commit_list(): don't free result\n - t5310: test delta reuse with bitmaps\n - bitmap_has_sha1_in_uninteresting(): drop BUG check\n (this branch uses jk/pack-delta-reuse-with-bitmap.)\n\n Hotfix of the base topic. It may not be a bad idea to squash these\n in when the next development cycle opens.\n\n\n* js/rebase-i-autosquash-fix (2018-09-04) 2 commits\n - rebase -i: be careful to wrap up fixup/squash chains\n - rebase -i --autosquash: demonstrate a problem skipping the last squash\n\n \"git rebase -i\" did not clear the state files correctly when a run\n of \"squash/fixup\" is aborted and then the user manually amended the\n commit instead, which has been corrected.\n\n Will merge to 'next'.\n\n\n* nd/bisect-show-list-fix (2018-09-04) 1 commit\n - bisect.c: make show_list() build again\n\n Debugging aid update.\n\n Will merge to and cook in 'next'.\n\n\n* nd/optim-reading-index-v4 (2018-09-04) 1 commit\n - read-cache.c: optimize reading index format v4\n\n The v4 format of the index file uses prefix compression to store\n the pathnames to save file size. The codepath to read such a file\n has been optimized.\n\n Will merge to and cook in 'next'.\n\n\n* sg/doc-trace-appends (2018-09-04) 1 commit\n - Documentation/git.txt: clarify that GIT_TRACE=/path appends\n\n Docfix.\n\n Will merge to 'next'.\n\n--------------------------------------------------\n[Stalled]\n\n* sl/commit-dry-run-with-short-output-fix (2018-07-30) 4 commits\n . commit: fix exit code when doing a dry run\n . wt-status: teach wt_status_collect about merges in progress\n . wt-status: rename commitable to committable\n . t7501: add coverage for flags which imply dry runs\n\n \"git commit --dry-run\" gave a correct exit status even during a\n conflict resolution toward a merge, but it did not with the\n \"--short\" option, which has been corrected.\n\n Seems to break 7512, 3404 and 7060 in 'pu'.\n\n\n* ma/wrapped-info (2018-05-28) 2 commits\n - usage: prefix all lines in `vreportf()`, not just the first\n - usage: extract `prefix_suffix_lines()` from `advise()`\n\n An attempt to help making multi-line messages fed to warning(),\n error(), and friends more easily translatable.\n\n Will discard and wait for a cleaned-up rewrite.\n cf. <20180529213957.GF7964@sigill.intra.peff.net>\n\n\n* hn/bisect-first-parent (2018-04-21) 1 commit\n - bisect: create 'bisect_flags' parameter in find_bisection()\n (this branch is used by tb/bisect-first-parent.)\n\n Preliminary code update to allow passing more flags down the\n bisection codepath in the future.\n\n We do not add random code that does not have real users to our\n codebase, so let's have it wait until such a real code materializes\n before too long.\n\n\n* av/fsmonitor-updates (2018-01-04) 6 commits\n - fsmonitor: use fsmonitor data in `git diff`\n - fsmonitor: remove debugging lines from t/t7519-status-fsmonitor.sh\n - fsmonitor: make output of test-dump-fsmonitor more concise\n - fsmonitor: update helper tool, now that flags are filled later\n - fsmonitor: stop inline'ing mark_fsmonitor_valid / _invalid\n - dir.c: update comments to match argument name\n\n Code clean-up on fsmonitor integration, plus optional utilization\n of the fsmonitor data in diff-files.\n\n Waiting for an update.\n cf. \n\n\n* pb/bisect-helper-2 (2018-07-23) 8 commits\n - t6030: make various test to pass GETTEXT_POISON tests\n - bisect--helper: `bisect_start` shell function partially in C\n - bisect--helper: `get_terms` & `bisect_terms` shell function in C\n - bisect--helper: `bisect_next_check` shell function in C\n - bisect--helper: `check_and_set_terms` shell function in C\n - wrapper: move is_empty_file() and rename it as is_empty_or_missing_file()\n - bisect--helper: `bisect_write` shell function in C\n - bisect--helper: `bisect_reset` shell function in C\n\n Expecting a reroll.\n cf. <0102015f5e5ee171-f30f4868-886f-47a1-a4e4-b4936afc545d-000000@eu-west-1.amazonses.com>\n\n I just rebased the topic to a newer base as it did not build\n standalone with the base I originally queued the topic on, but\n otherwise there is no update to address any of the review comments\n in the thread above---we are still waiting for a reroll.\n\n\n* jk/drop-ancient-curl (2017-08-09) 5 commits\n - http: #error on too-old curl\n - curl: remove ifdef'd code never used with curl >=7.19.4\n - http: drop support for curl < 7.19.4\n - http: drop support for curl < 7.16.0\n - http: drop support for curl < 7.11.1\n\n Some code in http.c that has bitrot is being removed.\n\n Expecting a reroll.\n\n\n* mk/use-size-t-in-zlib (2017-08-10) 1 commit\n . zlib.c: use size_t for size\n\n The wrapper to call into zlib followed our long tradition to use\n \"unsigned long\" for sizes of regions in memory, which have been\n updated to use \"size_t\".\n\n Needs resurrecting by making sure the fix is good and still applies\n (or adjusted to today's codebase).\n\n--------------------------------------------------\n[Cooking]\n\n* ds/format-commit-graph-docs (2018-08-21) 2 commits\n - commit-graph.txt: improve formatting for asciidoc\n - Docs: Add commit-graph tech docs to Makefile\n\n Design docs for the commit-graph machinery is now made into HTML as\n well as text.\n\n\n* jk/diff-rendered-docs (2018-08-31) 5 commits\n - doc/Makefile: drop doc-diff worktree and temporary files on \"make clean\"\n - doc-diff: add --clean mode to remove temporary working gunk\n - doc-diff: fix non-portable 'man' invocation\n - doc-diff: always use oids inside worktree\n (merged to 'next' on 2018-08-22 at dd7a2b71cd)\n + SubmittingPatches: mention doc-diff\n\n Dev doc update.\n\n Will merge to 'next'.\n\n\n* jk/pack-delta-reuse-with-bitmap (2018-08-21) 6 commits\n (merged to 'next' on 2018-08-22 at fc50b59dab)\n + pack-objects: reuse on-disk deltas for thin \"have\" objects\n + pack-bitmap: save \"have\" bitmap from walk\n + t/perf: add perf tests for fetches from a bitmapped server\n + t/perf: add infrastructure for measuring sizes\n + t/perf: factor out percent calculations\n + t/perf: factor boilerplate out of test_perf\n (this branch is used by jk/pack-objects-with-bitmap-fix.)\n\n When creating a thin pack, which allows objects to be made into a\n delta against another object that is not in the resulting pack but\n is known to be present on the receiving end, the code learned to\n take advantage of the reachability bitmap; this allows the server\n to send a delta against a base beyond the \"boundary\" commit.\n\n Will cook in 'next'.\n\n\n* jk/rev-list-stdin-noop-is-ok (2018-08-22) 1 commit\n (merged to 'next' on 2018-08-27 at d5916f7bc1)\n + rev-list: make empty --stdin not an error\n\n \"git rev-list --stdin for function exclusion\n\n spatch transformation to replace boolean uses of !hashcmp() to\n newly introduced oideq() is added, and applied, to regain\n performance lost due to support of multiple hash algorithms.\n\n Will cook in 'next'.\n\n\n* js/add-i-coalesce-after-editing-hunk (2018-08-28) 1 commit\n - add -p: coalesce hunks before testing applicability\n\n Applicability check after a patch is edited in a \"git add -i/p\"\n session has been improved.\n\n Will hold.\n cf. \n\n\n* rs/mailinfo-format-flowed (2018-08-29) 1 commit\n (merged to 'next' on 2018-08-31 at 9e8b92176c)\n + mailinfo: support format=flowed\n\n \"git mailinfo\" used in \"git am\" learned to make a best-effort\n recovery of a patch corrupted by MUA that sends text/plain with\n format=flawed option.\n\n Will cook in 'next'.\n\n\n* sb/submodule-move-head-with-corruption (2018-08-28) 2 commits\n - submodule.c: warn about missing submodule git directories\n - t2013: add test for missing but active submodule\n\n A corner case in switching to a branch that needs to \"check out\" a\n submodule is handled a bit better now.\n\n Expecting a reroll.\n\n\n* tg/conflict-marker-size (2018-08-29) 1 commit\n (merged to 'next' on 2018-08-31 at 12099161f0)\n + .gitattributes: add conflict-marker-size for relevant files\n\n Developer aid.\n\n Will cook in 'next'.\n\n\n* ts/doc-build-manpage-xsl-quietly (2018-08-29) 1 commit\n (merged to 'next' on 2018-08-31 at 7527e0f8d3)\n + Documentation/Makefile: make manpage-base-url.xsl generation quieter\n\n Build tweak.\n\n Will cook in 'next'.\n\n\n* bp/checkout-new-branch-optim (2018-08-16) 1 commit\n (merged to 'next' on 2018-08-27 at e69bfd115f)\n + checkout: optimize \"git checkout -b \"\n\n \"git checkout -b newbranch [HEAD]\" should not have to do as much as\n checking out a commit different from HEAD. An attempt is made to\n optimize this special case.\n\n Will cook in 'next'.\n\n\n* ao/submodule-wo-gitmodules-checked-out (2018-08-14) 7 commits\n - submodule: support reading .gitmodules even when it's not checked out\n - t7506: clean up .gitmodules properly before setting up new scenario\n - submodule: use the 'submodule--helper config' command\n - submodule--helper: add a new 'config' subcommand\n - t7411: be nicer to future tests and really clean things up\n - submodule: factor out a config_set_in_gitmodules_file_gently function\n - submodule: add a print_config_from_gitmodules() helper\n\n The submodule support has been updated to read from the blob at\n HEAD:.gitmodules when the .gitmodules file is missing from the\n working tree.\n\n Expecting a reroll.\n\n I find the design a bit iffy in that our usual \"missing in the\n working tree? let's use the latest blob\" fallback is to take it\n from the index, not from the HEAD.\n\n\n* bw/submodule-name-to-dir (2018-08-10) 2 commits\n (merged to 'next' on 2018-08-22 at c17f83be24)\n + submodule: munge paths to submodule git directories\n + submodule: create helper to build paths to submodule gitdirs\n\n In modern repository layout, the real body of a cloned submodule\n repository is held in .git/modules/ of the superproject, indexed by\n the submodule name. URLencode the submodule name before computing\n the name of the directory to make sure they form a flat namespace.\n\n Expecting further work on the topic.\n cf. \n\n\n* cc/delta-islands (2018-08-16) 7 commits\n (merged to 'next' on 2018-08-27 at cf3d7bd93f)\n + pack-objects: move 'layer' into 'struct packing_data'\n + pack-objects: move tree_depth into 'struct packing_data'\n + t5320: tests for delta islands\n + repack: add delta-islands support\n + pack-objects: add delta-islands support\n + pack-objects: refactor code into compute_layer_order()\n + Add delta-islands.{c,h}\n\n Lift code from GitHub to restrict delta computation so that an\n object that exists in one fork is not made into a delta against\n another object that does not appear in the same forked repository.\n\n Will cook in 'next'.\n\n\n* md/filter-trees (2018-09-04) 7 commits\n - list-objects-filter: implement filter tree:0\n - list-objects-filter: use BUG rather than die\n - revision: mark non-user-given objects instead\n - rev-list: handle missing tree objects properly\n - list-objects: always parse trees gently\n - list-objects: refactor to process_tree_contents\n - list-objects: store common func args in struct\n\n The \"rev-list --filter\" feature learned to exclude all trees via\n \"tree:0\" filter.\n\n Will merge to 'next'.\n\n\n* ng/status-i-short-for-ignored (2018-08-09) 1 commit\n - status: -i shorthand for --ignored command line option\n\n \"git status --ignored\" gained a shorthand \"git status -i\".\n\n What's the list opinion on this one? It is Meh to me, but\n obviously the author cared enough to write a patch, so...\n\n\n* pk/rebase-in-c-2-basic (2018-08-10) 11 commits\n - builtin rebase: support `git rebase `\n - builtin rebase: only store fully-qualified refs in `options.head_name`\n - builtin rebase: start a new rebase only if none is in progress\n - builtin rebase: support --force-rebase\n - builtin rebase: try to fast forward when possible\n - builtin rebase: require a clean worktree\n - builtin rebase: support the `verbose` and `diffstat` options\n - builtin rebase: support --quiet\n - builtin rebase: handle the pre-rebase hook (and add --no-verify)\n - builtin rebase: support `git rebase --onto A...B`\n - builtin rebase: support --onto\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final; uses pk/rebase-in-c.)\n\n\n* pk/rebase-in-c-3-acts (2018-08-10) 7 commits\n - builtin rebase: stop if `git am` is in progress\n - builtin rebase: actions require a rebase in progress\n - builtin rebase: support --edit-todo and --show-current-patch\n - builtin rebase: support --quit\n - builtin rebase: support --abort\n - builtin rebase: support --skip\n - builtin rebase: support --continue\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final; uses pk/rebase-in-c and pk/rebase-in-c-2-basic.)\n\n\n* pk/rebase-in-c-4-opts (2018-08-10) 18 commits\n - builtin rebase: support --root\n - builtin rebase: add support for custom merge strategies\n - builtin rebase: support `fork-point` option\n - merge-base --fork-point: extract libified function\n - builtin rebase: support --rebase-merges[=[no-]rebase-cousins]\n - builtin rebase: support `--allow-empty-message` option\n - builtin rebase: support `--exec`\n - builtin rebase: support `--autostash` option\n - builtin rebase: support `-C` and `--whitespace=`\n - builtin rebase: support `--gpg-sign` option\n - builtin rebase: support `--autosquash`\n - builtin rebase: support `keep-empty` option\n - builtin rebase: support `ignore-date` option\n - builtin rebase: support `ignore-whitespace` option\n - builtin rebase: support --committer-date-is-author-date\n - builtin rebase: support --rerere-autoupdate\n - builtin rebase: support --signoff\n - builtin rebase: allow selecting the rebase \"backend\"\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final; uses pk/rebase-in-c, pk/rebase-in-c-2-basic and pk/rebase-in-c-3-acts.)\n\n\n* pk/rebase-in-c-5-test (2018-08-10) 6 commits\n - builtin rebase: error out on incompatible option/mode combinations\n - builtin rebase: use no-op editor when interactive is \"implied\"\n - builtin rebase: show progress when connected to a terminal\n - builtin rebase: fast-forward to onto if it is a proper descendant\n - builtin rebase: optionally pass custom reflogs to reset_head()\n - builtin rebase: optionally auto-detect the upstream\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c and pk/rebase-in-c-6-final; uses pk/rebase-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts and pk/rebase-in-c-4-opts.)\n\n\n* pk/rebase-in-c-6-final (2018-08-29) 1 commit\n - rebase: default to using the builtin rebase\n (this branch is used by jc/rebase-in-c-9-fixes; uses ag/rebase-i-in-c, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts and pk/rebase-in-c-5-test.)\n\n\n* ps/stash-in-c (2018-08-31) 20 commits\n - stash: replace all `write-tree` child processes with API calls\n - stash: optimize `get_untracked_files()` and `check_changes()`\n - stash: convert `stash--helper.c` into `stash.c`\n - stash: convert save to builtin\n - stash: make push -q quiet\n - stash: convert push to builtin\n - stash: convert create to builtin\n - stash: convert store to builtin\n - stash: mention options in `show` synopsis\n - stash: convert show to builtin\n - stash: convert list to builtin\n - stash: convert pop to builtin\n - stash: convert branch to builtin\n - stash: convert drop and clear to builtin\n - stash: convert apply to builtin\n - stash: add tests for `git stash show` config\n - stash: rename test cases to be more descriptive\n - stash: update test cases conform to coding guidelines\n - stash: improve option parsing test coverage\n - sha1-name.c: add `get_oidf()` which acts like `get_oid()`\n\n\n* nd/clone-case-smashing-warning (2018-08-17) 1 commit\n (merged to 'next' on 2018-08-22 at eedae40a8d)\n + clone: report duplicate entries on case-insensitive filesystems\n\n Running \"git clone\" against a project that contain two files with\n pathnames that differ only in cases on a case insensitive\n filesystem would result in one of the files lost because the\n underlying filesystem is incapable of holding both at the same\n time. An attempt is made to detect such a case and warn.\n\n Will cook in 'next'.\n\n\n* nd/unpack-trees-with-cache-tree (2018-08-27) 8 commits\n (merged to 'next' on 2018-08-27 at b1d841d034)\n + Document update for nd/unpack-trees-with-cache-tree\n (merged to 'next' on 2018-08-22 at 138b902673)\n + cache-tree: verify valid cache-tree in the test suite\n + unpack-trees: add missing cache invalidation\n + unpack-trees: reuse (still valid) cache-tree from src_index\n + unpack-trees: reduce malloc in cache-tree walk\n + unpack-trees: optimize walking same trees with cache-tree\n + unpack-trees: add performance tracing\n + trace.h: support nested performance tracing\n\n The unpack_trees() API used in checking out a branch and merging\n walks one or more trees along with the index. When the cache-tree\n in the index tells us that we are walking a tree whose flattened\n contents is known (i.e. matches a span in the index), as linearly\n scanning a span in the index is much more efficient than having to\n open tree objects recursively and listing their entries, the walk\n can be optimized, which is done in this topic.\n\n Will cook in 'next'.\n\n\n* sb/range-diff-colors (2018-08-20) 11 commits\n (merged to 'next' on 2018-08-22 at eb7ed4fca3)\n + range-diff: indent special lines as context\n + range-diff: make use of different output indicators\n + diff.c: add --output-indicator-{new, old, context}\n + diff.c: rewrite emit_line_0 more understandably\n + diff.c: omit check for line prefix in emit_line_0\n + diff: use emit_line_0 once per line\n + diff.c: add set_sign to emit_line_0\n + diff.c: reorder arguments for emit_line_ws_markup\n + diff.c: simplify caller of emit_line_0\n + t3206: add color test for range-diff --dual-color\n + test_decode_color: understand FAINT and ITALIC\n\n The color output support for recently introduced \"range-diff\"\n command got tweaked a bit.\n\n Will cook in 'next'.\n\n\n* sg/t1404-update-ref-test-timeout (2018-08-01) 1 commit\n (merged to 'next' on 2018-08-22 at f3cd37b5ea)\n + t1404: increase core.packedRefsTimeout to avoid occasional test failure\n\n An attempt to unflake a test a bit.\n\n Will cook in 'next'.\n\n\n* pw/add-p-select (2018-07-26) 4 commits\n - add -p: optimize line selection for short hunks\n - add -p: allow line selection to be inverted\n - add -p: select modified lines correctly\n - add -p: select individual hunk lines\n\n \"git add -p\" interactive interface learned to let users choose\n individual added/removed lines to be used in the operation, instead\n of accepting or rejecting a whole hunk.\n\n Will hold.\n cf. \n I found the feature to be hard to explain, and may result in more\n end-user complaints, but let's see.\n\n\n* ds/commit-graph-with-grafts (2018-08-21) 8 commits\n - commit-graph: close_commit_graph before shallow walk\n - commit-graph: not compatible with uninitialized repo\n - commit-graph: not compatible with grafts\n - commit-graph: not compatible with replace objects\n - test-repository: properly init repo\n - commit-graph: update design document\n - refs.c: upgrade for_each_replace_ref to be a each_repo_ref_fn callback\n - refs.c: migrate internal ref iteration to pass thru repository argument\n\n The recently introduced commit-graph auxiliary data is incompatible\n with mechanisms such as replace & grafts that \"breaks\" immutable\n nature of the object reference relationship. Disable optimizations\n based on its use (and updating existing commit-graph) when these\n incompatible features are in use in the repository.\n\n Replaced with a newer version.\n\n\n* ds/reachable (2018-08-28) 19 commits\n (merged to 'next' on 2018-08-28 at b1634b371d)\n + commit-reach: correct accidental #include of C file\n (merged to 'next' on 2018-08-22 at 17f3275afb)\n + commit-reach: use can_all_from_reach\n + commit-reach: make can_all_from_reach... linear\n + commit-reach: replace ref_newer logic\n + test-reach: test commit_contains\n + test-reach: test can_all_from_reach_with_flags\n + test-reach: test reduce_heads\n + test-reach: test get_merge_bases_many\n + test-reach: test is_descendant_of\n + test-reach: test in_merge_bases\n + test-reach: create new test tool for ref_newer\n + commit-reach: move can_all_from_reach_with_flags\n + upload-pack: generalize commit date cutoff\n + upload-pack: refactor ok_to_give_up()\n + upload-pack: make reachable() more generic\n + commit-reach: move commit_contains from ref-filter\n + commit-reach: move ref_newer from remote.c\n + commit.h: remove method declarations\n + commit-reach: move walk methods from commit.c\n\n The code for computing history reachability has been shuffled,\n obtained a bunch of new tests to cover them, and then being\n improved.\n\n Will cook in 'next'.\n\n\n* es/format-patch-interdiff (2018-07-23) 6 commits\n (merged to 'next' on 2018-08-31 at 63927e0227)\n + format-patch: allow --interdiff to apply to a lone-patch\n + log-tree: show_log: make commentary block delimiting reusable\n + interdiff: teach show_interdiff() to indent interdiff\n + format-patch: teach --interdiff to respect -v/--reroll-count\n + format-patch: add --interdiff option to embed diff in cover letter\n + format-patch: allow additional generated content in make_cover_letter()\n (this branch is used by es/format-patch-rangediff.)\n\n \"git format-patch\" learned a new \"--interdiff\" option to explain\n the difference between this version and the previous atttempt in\n the cover letter (or after the tree-dashes as a comment).\n\n Will cook in 'next'.\n\n\n* es/format-patch-rangediff (2018-08-14) 10 commits\n (merged to 'next' on 2018-08-31 at 65627afece)\n + format-patch: allow --range-diff to apply to a lone-patch\n + format-patch: add --creation-factor tweak for --range-diff\n + format-patch: teach --range-diff to respect -v/--reroll-count\n + format-patch: extend --range-diff to accept revision range\n + format-patch: add --range-diff option to embed diff in cover letter\n + range-diff: relieve callers of low-level configuration burden\n + range-diff: publish default creation factor\n + range-diff: respect diff_option.file rather than assuming 'stdout'\n + Merge branch 'es/format-patch-interdiff' into es/format-patch-rangediff\n + Merge branch 'js/range-diff' into es/format-patch-rangediff\n (this branch uses es/format-patch-interdiff.)\n\n \"git format-patch\" learned a new \"--range-diff\" option to explain\n the difference between this version and the previous attempt in\n the cover letter (or after the tree-dashes as a comment).\n\n Will cook in 'next'.\n\n\n* jn/gc-auto (2018-07-17) 3 commits\n - gc: do not return error for prior errors in daemonized mode\n - gc: exit with status 128 on failure\n - gc: improve handling of errors reading gc.log\n\n \"gc --auto\" ended up calling exit(-1) upon error, which has been\n corrected to use exit(1). Also the error reporting behaviour when\n daemonized has been updated to exit with zero status when stopping\n due to a previously discovered error (which implies there is no\n point running gc to improve the situation); we used to exit with\n failure in such a case.\n\n What's the donness of this one?\n cf. <20180717201348.GD26218@sigill.intra.peff.net>\n\n\n* sb/submodule-update-in-c (2018-08-14) 7 commits\n (merged to 'next' on 2018-08-17 at 23c81e5ff7)\n + submodule--helper: introduce new update-module-mode helper\n + submodule--helper: replace connect-gitdir-workingtree by ensure-core-worktree\n + builtin/submodule--helper: factor out method to update a single submodule\n + builtin/submodule--helper: store update_clone information in a struct\n + builtin/submodule--helper: factor out submodule updating\n + git-submodule.sh: rename unused variables\n + git-submodule.sh: align error reporting for update mode to use path\n\n \"git submodule update\" is getting rewritten piece-by-piece into C.\n\n Will cook in 'next'.\n\n\n* tg/rerere (2018-08-06) 11 commits\n (merged to 'next' on 2018-08-17 at 919a958cdc)\n + rerere: recalculate conflict ID when unresolved conflict is committed\n + rerere: teach rerere to handle nested conflicts\n + rerere: return strbuf from handle path\n + rerere: factor out handle_conflict function\n + rerere: only return whether a path has conflicts or not\n + rerere: fix crash with files rerere can't handle\n + rerere: add documentation for conflict normalization\n + rerere: mark strings for translation\n + rerere: wrap paths in output in sq\n + rerere: lowercase error messages\n + rerere: unify error messages when read_cache fails\n (this branch is used by tg/rerere-doc-updates.)\n\n Fixes to \"git rerere\" corner cases, especially when conflict\n markers cannot be parsed in the file.\n\n Will cook in 'next'.\n\n\n* ag/rebase-i-in-c (2018-08-29) 20 commits\n - rebase -i: move rebase--helper modes to rebase--interactive\n - rebase -i: remove git-rebase--interactive.sh\n - rebase--interactive2: rewrite the submodes of interactive rebase in C\n - rebase -i: implement the main part of interactive rebase as a builtin\n - rebase -i: rewrite init_basic_state() in C\n - rebase -i: rewrite write_basic_state() in C\n - rebase -i: rewrite the rest of init_revisions_and_shortrevisions() in C\n - rebase -i: implement the logic to initialize $revisions in C\n - rebase -i: remove unused modes and functions\n - rebase -i: rewrite complete_action() in C\n - t3404: todo list with commented-out commands only aborts\n - sequencer: change the way skip_unnecessary_picks() returns its result\n - sequencer: refactor append_todo_help() to write its message to a buffer\n - rebase -i: rewrite checkout_onto() in C\n - rebase -i: rewrite setup_reflog_action() in C\n - sequencer: add a new function to silence a command, except if it fails\n - rebase -i: rewrite the edit-todo functionality in C\n - editor: add a function to launch the sequence editor\n - rebase -i: rewrite append_todo_help() in C\n - sequencer: make three functions and an enum from sequencer.c public\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c and pk/rebase-in-c-6-final.)\n\n Rewrite of the remaining \"rebase -i\" machinery in C.\n\n\n* lt/date-human (2018-07-09) 1 commit\n - Add 'human' date format\n\n A new date format \"--date=human\" that morphs its output depending\n on how far the time is from the current time has been introduced.\n \"--date=auto\" can be used to use this new format when the output is\n goint to the pager or to the terminal and otherwise the default\n format.\n\n\n* pk/rebase-in-c (2018-08-06) 3 commits\n - builtin/rebase: support running \"git rebase \"\n - rebase: refactor common shell functions into their own file\n - rebase: start implementing it as a builtin\n (this branch is used by jc/rebase-in-c-9-fixes, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final.)\n\n Rewrite of the \"rebase\" machinery in C.\n\n\n* jk/branch-l-1-repurpose (2018-08-30) 2 commits\n (merged to 'next' on 2018-08-31 at cfa73bbfcb)\n + doc/git-branch: remove obsolete \"-l\" references\n (merged to 'next' on 2018-08-08 at d2a08dd08e)\n + branch: make \"-l\" a synonym for \"--list\"\n\n Updated plan to repurpose the \"-l\" option to \"git branch\".\n\n Will cook in 'next'.\n\n\n* ds/multi-pack-index (2018-08-20) 33 commits\n (merged to 'next' on 2018-08-21 at d15e8cadd4)\n + pack-objects: consider packs in multi-pack-index\n + midx: test a few commands that use get_all_packs\n + treewide: use get_all_packs\n + packfile: add all_packs list\n + midx: fix bug that skips midx with alternates\n + midx: stop reporting garbage\n + midx: mark bad packed objects\n + multi-pack-index: store local property\n + multi-pack-index: provide more helpful usage info\n + Sync 'ds/multi-pack-index' to v2.19.0-rc0\n (merged to 'next' on 2018-08-08 at 1a56c52967)\n + midx: clear midx on repack\n + packfile: skip loading index if in multi-pack-index\n + midx: prevent duplicate packfile loads\n + midx: use midx in approximate_object_count\n + midx: use existing midx when writing new one\n + midx: use midx in abbreviation calculations\n + midx: read objects from multi-pack-index\n + config: create core.multiPackIndex setting\n + midx: write object offsets\n + midx: write object id fanout chunk\n + midx: write object ids in a chunk\n + midx: sort and deduplicate objects from packfiles\n + midx: read pack names into array\n + multi-pack-index: write pack names in chunk\n + multi-pack-index: read packfile list\n + packfile: generalize pack directory list\n + t5319: expand test data\n + multi-pack-index: load into memory\n + midx: write header information to lockfile\n + multi-pack-index: add 'write' verb\n + multi-pack-index: add builtin\n + multi-pack-index: add format details\n + multi-pack-index: add design document\n\n When there are too many packfiles in a repository (which is not\n recommended), looking up an object in these would require\n consulting many pack .idx files; a new mechanism to have a single\n file that consolidates all of these .idx files is introduced.\n\n Will cook in 'next'.\n\n--------------------------------------------------\n[Discarded]\n\n* am/sequencer-author-script-fix (2018-07-18) 1 commit\n . sequencer.c: terminate the last line of author-script properly\n\n The author-script that records the author information created by\n the sequencer machinery lacked the closing single quote on the last\n entry.\n\n Superseded by another topic.\n\n\n* jc/push-cas-opt-comment (2018-08-01) 1 commit\n . push: comment on a funny unbalanced option help\n\n Code clarification.\n\n Superseded by another topic.\n\n\n* cc/remote-odb (2018-08-02) 9 commits\n . Documentation/config: add odb..promisorRemote\n . t0410: test fetching from many promisor remotes\n . Use odb.origin.partialclonefilter instead of core.partialclonefilter\n . Use remote_odb_get_direct() and has_remote_odb()\n . remote-odb: add remote_odb_reinit()\n . remote-odb: implement remote_odb_get_many_direct()\n . remote-odb: implement remote_odb_get_direct()\n . Add initial remote odb support\n . fetch-object: make functions return an error code\n\n Implement lazy fetches of missing objects to complement the\n experimental partial clone feature.\n\n Ejected; seems to break existing repositories that use partialclone\n repository extension.\n\n\n* jh/structured-logging (2018-08-28) 26 commits\n . SQUASH??? spatch fix\n . structured-logging: add config data facility\n . structured-logging: t0420 tests for interacitve child_summary\n . structured-logging: t0420 tests for child process detail events\n . structured-logging: add child process classification\n . structured-logging: add detail-events for child processes\n . structured-logging: add structured logging to remote-curl\n . structured-logging: t0420 tests for aux-data\n . structured-logging: add aux-data for size of sparse-checkout file\n . structured-logging: add aux-data for index size\n . structured-logging: add aux-data facility\n . structured-logging: t0420 tests for timers\n . structured-logging: add timer around preload_index\n . structured-logging: add timer around wt-status functions\n . structured-logging: add timer around do_write_index\n . structured-logging: add timer around do_read_index\n . structured-logging: add timer facility\n . structured-logging: add detail-event for lazy_init_name_hash\n . structured-logging: add detail-event facility\n . structured-logging: t0420 basic tests\n . structured-logging: set sub_command field for checkout command\n . structured-logging: set sub_command field for branch command\n . structured-logging: add session-id to log events\n . structured-logging: add structured logging framework\n . structured-logging: add STRUCTURED_LOGGING=1 to Makefile\n . structured-logging: design document\n\n Being rerolled with an updated tracing API.\n"},{"id":"357408","messageId":"87bm9cs5y6.fsf@evledraar.gmail.com","threadId":"49277","inReplyTo":"xmqqmusw6gbo.fsf@gitster-ct.c.googlers.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Ævar Arnfjörð Bjarmason","fromEmail":"avarab@gmail.com","sentAt":"2018-09-05T08:29:21Z","receivedAt":"2018-09-05T08:29:26Z","isPatch":false,"sender":{"key":"avarab@gmail.com","avatar":"https://avatars.githubusercontent.com/u/45301?v=4"},"body":"\nOn Tue, Sep 04 2018, Junio C Hamano wrote:\n\n> Git 2.19-rc2 is out. Hopefully the tip of 'master' is more or less\n> identical to the final one without needing much updates.\n>\n> You can find the changes described here in the integration branches\n> of the repositories listed at\n>\n> http://git-blame.blogspot.com/p/git-public-repositories.html\n>\n> --------------------------------------------------\n> [Graduated to \"master\"]\n>\n> * ab/portable (2018-08-27) 6 commits\n> (merged to 'next' on 2018-08-27 at 37640e66ef)\n> + tests: fix and add lint for non-portable grep --file\n> + tests: fix version-specific portability issue in Perl JSON\n> + tests: use shorter labels in chainlint.sed for AIX sed\n> + tests: fix comment syntax in chainlint.sed for AIX sed\n> + tests: fix and add lint for non-portable seq\n> + tests: fix and add lint for non-portable head -c N\n> (this branch is used by ab/portable-more.)\n\nI recently gained access to a Solaris 10 SPARC (5.10) box and discovered\nthat the chainlint.sed implementation in 2.19.0 has more sed portability\nissues.\n\nFirst, whoever implemented the /bin/sed on Solaris apparently read the\nPOSIX requirements for a max length label of 8 to mean that 8 characters\nshould include the colon, so a bunch of things fail because of that, but\nare fixed with a shorter 7 character label.\n\nThen GIT_TEST_CHAIN_LINT=1 still fails because 878f988350 (\"t/test-lib:\nteach --chain-lint to detect broken &&-chains in subshells\", 2018-07-11)\nadded a \"grep -q\" invocation. The /bin/grep on that version of Solaris\ndoesn't have -q. We fixed a similar issue way back in 80700fde91\n(\"t/t1304: make a second colon optional in the mask ACL check\",\n2010-03-15) by redirecting to /dev/null instead.\n\nA bunch of other tests in the test suite rely on \"grep -q\", but nothing\nas central as chainlint, so it makes everything break. Do we want to\naway with \"grep -q\" entirely because of old Solaris /bin/grep?\n\nAt this point those familiar with Solaris are screaming (\"why are you\nusing anything in /bin!\"). Okey fine, but it mostly worked before, so\nare we OK with breaking it? \"Mostly\" here is \"test suite would fail\n20-30 tests for various reasons, but at least no failures in test-lib.sh\nand the like\".\n\nHowever, if as config.mak.uname does we run the tests with\nPATH=/usr/xpg6/bin:/usr/xpg4/bin:$PATH, at this point sed is fine with 8\ncharacter labels, but starts complaining about this (also in\nchainlint.sed):\n\n sed: Too many commands, last: s/\\n//\n\nAs with other sed issues I fixed recently in chainlint.sed this one is\njust the tip of the iceberg. Once you \"fix\" one (just remove it, I have\nno idea how to rewrite it) others appear.\n\nI was hoping this would just be a Solaris 10 issue, but it's also an\nissue in Solaris 11 (5.11 11.3).\n\nSo, do we want to chase this down or just do this?:\n\n diff --git a/Makefile b/Makefile\n index 5a969f5..f125dc5 100644\n --- a/Makefile\n +++ b/Makefile\n @@ -2602,6 +2602,9 @@ endif\n ifdef TEST_GIT_INDEX_VERSION\n @echo TEST_GIT_INDEX_VERSION=\\''$(subst ','\\'',$(subst ','\\'',$(TEST_GIT_INDEX_VERSION)))'\\' >>$@+\n endif\n +ifdef GIT_TEST_CHAIN_LINT\n + @echo GIT_TEST_CHAIN_LINT=\\''$(subst ','\\'',$(subst ','\\'',$(GIT_TEST_CHAIN_LINT)))'\\' >>$@+\n +endif\n @if cmp $@+ $@ >/dev/null 2>&1; then $(RM) $@+; else mv $@+ $@; fi\n\n ### Detect Python interpreter path changes\n diff --git a/config.mak.uname b/config.mak.uname\n index e47af72..2b02a2b 100644\n --- a/config.mak.uname\n +++ b/config.mak.uname\n @@ -163,6 +163,10 @@ ifeq ($(uname_S),SunOS)\n INSTALL = /usr/ucb/install\n TAR = gtar\n BASIC_CFLAGS += -D__EXTENSIONS__ -D__sun__\n + # t/chainlint.sed is hopelessly broken all known (tested\n + # Solaris 10 & 11) versions of Solaris, both /bin/sed and\n + # /usr/xpg4/bin/sed\n + GIT_TEST_CHAIN_LINT = 0\n endif\n ifeq ($(uname_O),Cygwin)\n ifeq ($(shell expr \"$(uname_R)\" : '1\\.[1-6]\\.'),4)\n\nIf I could wave a magic wand I'd say \"let's just rewrite chainlint.sed\nin perl, at least that's portable\", but that's a lot of effort (and I\ndoubt Eric wants to). It slightly sucks to not have chainlint on\nSolaris, but it would also suck to revert chainlint.sed back to 2.18.0\n(there were some big improvements). So I think the patch above is the\nbest way forward, especially since we're on rc2. What do you think?\n"},{"id":"357410","messageId":"CAPig+cRrwW662Hzp8V1sH51x8qYxPyqG=VHpOkfd-=RwUSSZbw@mail.gmail.com","threadId":"49277","inReplyTo":"87bm9cs5y6.fsf@evledraar.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Eric Sunshine","fromEmail":"sunshine@sunshineco.com","sentAt":"2018-09-05T08:59:27Z","receivedAt":"2018-09-05T08:59:42Z","isPatch":false,"sender":{"key":"sunshine@sunshineco.com","avatar":"https://avatars.githubusercontent.com/u/163641?v=4"},"body":"On Wed, Sep 5, 2018 at 4:29 AM Ævar Arnfjörð Bjarmason wrote:\n> I recently gained access to a Solaris 10 SPARC (5.10) box and discovered\n> that the chainlint.sed implementation in 2.19.0 has more sed portability\n> issues.\n>\n> First, whoever implemented the /bin/sed on Solaris apparently read the\n> POSIX requirements for a max length label of 8 to mean that 8 characters\n> should include the colon, so a bunch of things fail because of that, but\n> are fixed with a shorter 7 character label.\n\nI'm pretty sure that Solaris 'sed' predates POSIX by a good bit, but\nthat's neither here nor there.\n\n> Then GIT_TEST_CHAIN_LINT=1 still fails because 878f988350 (\"t/test-lib:\n> teach --chain-lint to detect broken &&-chains in subshells\", 2018-07-11)\n> added a \"grep -q\" invocation. The /bin/grep on that version of Solaris\n> doesn't have -q.\n\nI knew that '-q' was potentially problematic on some platforms, so I\nchecked and saw that existing tests were already using it, thus went\nahead and used it. Dropping '-q' here and redirecting stderr to\n/dev/null is a perfectly fine alternative.\n\n> We fixed a similar issue way back in 80700fde91\n> (\"t/t1304: make a second colon optional in the mask ACL check\",\n> 2010-03-15) by redirecting to /dev/null instead.\n>\n> A bunch of other tests in the test suite rely on \"grep -q\", but nothing\n> as central as chainlint, so it makes everything break. Do we want to\n> away with \"grep -q\" entirely because of old Solaris /bin/grep?\n\nI count 132 instances in existing tests (though, I may have missed some).\n\n> At this point those familiar with Solaris are screaming (\"why are you\n> using anything in /bin!\"). Okey fine, but it mostly worked before, so\n> are we OK with breaking it? \"Mostly\" here is \"test suite would fail\n> 20-30 tests for various reasons, but at least no failures in test-lib.sh\n> and the like\".\n>\n> However, if as config.mak.uname does we run the tests with\n> PATH=/usr/xpg6/bin:/usr/xpg4/bin:$PATH, at this point sed is fine with 8\n> character labels [...]\n\nSo, if you run the tests via 'make test' (or whatever), then you get\n/usr/xpg?/bin in PATH, but if you run an individual test script (and\nhaven't set your PATH appropriately), then you encounter the problems\nyou report above?\n\n> [...] but starts complaining about this (also in\n> chainlint.sed):\n>\n> sed: Too many commands, last: s/\\n//\n\nErm. Any additional context to help narrow this down?\n\n> diff --git a/config.mak.uname b/config.mak.uname\n> @@ -163,6 +163,10 @@ ifeq ($(uname_S),SunOS)\n> BASIC_CFLAGS += -D__EXTENSIONS__ -D__sun__\n> + # t/chainlint.sed is hopelessly broken all known (tested\n> + # Solaris 10 & 11) versions of Solaris, both /bin/sed and\n> + # /usr/xpg4/bin/sed\n> + GIT_TEST_CHAIN_LINT = 0\n> endif\n>\n> It slightly sucks to not have chainlint on\n> Solaris, but it would also suck to revert chainlint.sed back to 2.18.0\n> (there were some big improvements). So I think the patch above is the\n> best way forward, especially since we're on rc2. What do you think?\n\nKeeping in mind that the main goal of 'chainlint' is to prevent _new_\nbreakage from entering the test suite, disabling 'chainlint' on\nSolaris is an entirely reasonable way forward. In present day, it\nseems quite unlikely that we'll see someone develop new tests on\nSolaris, so having 'chainlint' disabled there isn't likely to be a big\ndeal. Moreover, if someone does write a new test on Solaris which has\na broken &&-chain in a subshell, that breakage will be caught very\nquickly once the test is run by anyone on Linux or MacOS (or Windows\nor BSD or AIX), so it's not like the broken test will make it into the\nproject undetected.\n"},{"id":"357411","messageId":"CAPig+cTQLED6P2AWrxU=pvNc6WzO-5YVsPNh3fg28LdZqU15AA@mail.gmail.com","threadId":"49277","inReplyTo":"xmqqmusw6gbo.fsf@gitster-ct.c.googlers.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Eric Sunshine","fromEmail":"sunshine@sunshineco.com","sentAt":"2018-09-05T09:14:13Z","receivedAt":"2018-09-05T09:14:26Z","isPatch":false,"sender":{"key":"sunshine@sunshineco.com","avatar":"https://avatars.githubusercontent.com/u/163641?v=4"},"body":"On Tue, Sep 4, 2018 at 6:36 PM Junio C Hamano wrote:\n> * es/worktree-forced-ops-fix (2018-08-30) 9 commits\n> - worktree: delete .git/worktrees if empty after 'remove'\n> - worktree: teach 'remove' to override lock when --force given twice\n> - worktree: teach 'move' to override lock when --force given twice\n> - worktree: teach 'add' to respect --force for registered but missing path\n> - worktree: disallow adding same path multiple times\n> - worktree: prepare for more checks of whether path can become worktree\n> - worktree: generalize delete_git_dir() to reduce code duplication\n> - worktree: move delete_git_dir() earlier in file for upcoming new callers\n> - worktree: don't die() in library function find_worktree()\n>\n> Various subcommands of \"git worktree\" take '--force' but did not\n> behave sensibly, which has been corrected.\n\nThis description mischaracterizes what these changes are about. The\nprimary intent of this series is to fix a bug in which the same path\ncan be registered under multiple worktree entries.\n\nAs for --force, it worked perfectly fine for the couple git-worktree\nsubcommands which accepted it. The patches in this series dealing\nwith --force are merely extending it to other subcommands or to other\nuse-cases.\n\nSo, perhaps rewrite this description like this:\n\n Fix a bug in which the same path could be registered under\n multiple worktree entries if the patch was missing (for instance,\n was removed manually).\n\n Also, as a convenience, expand the number of cases in which\n --force is applicable.\n"},{"id":"357412","messageId":"CAPig+cSETNc5nRxf_zQq6JE8zNyUvKmdPWNsm6=NMSOXm2DV=A@mail.gmail.com","threadId":"49277","inReplyTo":"xmqqmusw6gbo.fsf@gitster-ct.c.googlers.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Eric Sunshine","fromEmail":"sunshine@sunshineco.com","sentAt":"2018-09-05T09:50:04Z","receivedAt":"2018-09-05T09:50:17Z","isPatch":false,"sender":{"key":"sunshine@sunshineco.com","avatar":"https://avatars.githubusercontent.com/u/163641?v=4"},"body":"On Tue, Sep 4, 2018 at 6:36 PM Junio C Hamano wrote:\n> * es/worktree-forced-ops-fix (2018-08-30) 9 commits\n> - worktree: delete .git/worktrees if empty after 'remove'\n> - worktree: teach 'remove' to override lock when --force given twice\n> - worktree: teach 'move' to override lock when --force given twice\n> - worktree: teach 'add' to respect --force for registered but missing path\n> - worktree: disallow adding same path multiple times\n> - worktree: prepare for more checks of whether path can become worktree\n> - worktree: generalize delete_git_dir() to reduce code duplication\n> - worktree: move delete_git_dir() earlier in file for upcoming new callers\n> - worktree: don't die() in library function find_worktree()\n>\n> Various subcommands of \"git worktree\" take '--force' but did not\n> behave sensibly, which has been corrected.\n\nThis series is missing a patch[1] from Peff which he wanted placed at\nthe end of the series. It was probably overlooked since he embedded\nit as a reply in that thread rather than sending it as a standalone\npatch.\n\n[1]: https://public-inbox.org/git/20180830075431.GF11944@sigill.intra.peff.net/\n"},{"id":"357413","messageId":"877ek0rymz.fsf@evledraar.gmail.com","threadId":"49277","inReplyTo":"CAPig+cRrwW662Hzp8V1sH51x8qYxPyqG=VHpOkfd-=RwUSSZbw@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Ævar Arnfjörð Bjarmason","fromEmail":"avarab@gmail.com","sentAt":"2018-09-05T11:07:16Z","receivedAt":"2018-09-05T11:07:23Z","isPatch":false,"sender":{"key":"avarab@gmail.com","avatar":"https://avatars.githubusercontent.com/u/45301?v=4"},"body":"\nOn Wed, Sep 05 2018, Eric Sunshine wrote:\n\n> On Wed, Sep 5, 2018 at 4:29 AM Ævar Arnfjörð Bjarmason wrote:\n>> I recently gained access to a Solaris 10 SPARC (5.10) box and discovered\n>> that the chainlint.sed implementation in 2.19.0 has more sed portability\n>> issues.\n>>\n>> First, whoever implemented the /bin/sed on Solaris apparently read the\n>> POSIX requirements for a max length label of 8 to mean that 8 characters\n>> should include the colon, so a bunch of things fail because of that, but\n>> are fixed with a shorter 7 character label.\n>\n> I'm pretty sure that Solaris 'sed' predates POSIX by a good bit, but\n> that's neither here nor there.\n>\n>> Then GIT_TEST_CHAIN_LINT=1 still fails because 878f988350 (\"t/test-lib:\n>> teach --chain-lint to detect broken &&-chains in subshells\", 2018-07-11)\n>> added a \"grep -q\" invocation. The /bin/grep on that version of Solaris\n>> doesn't have -q.\n>\n> I knew that '-q' was potentially problematic on some platforms, so I\n> checked and saw that existing tests were already using it, thus went\n> ahead and used it. Dropping '-q' here and redirecting stderr to\n> /dev/null is a perfectly fine alternative.\n>\n>> We fixed a similar issue way back in 80700fde91\n>> (\"t/t1304: make a second colon optional in the mask ACL check\",\n>> 2010-03-15) by redirecting to /dev/null instead.\n>>\n>> A bunch of other tests in the test suite rely on \"grep -q\", but nothing\n>> as central as chainlint, so it makes everything break. Do we want to\n>> away with \"grep -q\" entirely because of old Solaris /bin/grep?\n>\n> I count 132 instances in existing tests (though, I may have missed some).\n>\n>> At this point those familiar with Solaris are screaming (\"why are you\n>> using anything in /bin!\"). Okey fine, but it mostly worked before, so\n>> are we OK with breaking it? \"Mostly\" here is \"test suite would fail\n>> 20-30 tests for various reasons, but at least no failures in test-lib.sh\n>> and the like\".\n>>\n>> However, if as config.mak.uname does we run the tests with\n>> PATH=/usr/xpg6/bin:/usr/xpg4/bin:$PATH, at this point sed is fine with 8\n>> character labels [...]\n>\n> So, if you run the tests via 'make test' (or whatever), then you get\n> /usr/xpg?/bin in PATH, but if you run an individual test script (and\n> haven't set your PATH appropriately), then you encounter the problems\n> you report above?\n\nYou need to manually set the PATH before running the tests, the\nSANE_TOOL_PATH just affects git-sh-setup. We should probably fix that,\ni.e. now if you compile git our shellscripts will use the fixed paths,\nbut not the tests.\n\n>> [...] but starts complaining about this (also in\n>> chainlint.sed):\n>>\n>> sed: Too many commands, last: s/\\n//\n>\n> Erm. Any additional context to help narrow this down?\n\nI tried playing around with it a bit, but honestly don't understand\nenough of this advanced sed syntax to make any headway, it's complaining\nabout e.g. the \"check for multi-line double-quoted string\" rule, but\nthen if you remove the s/\\n// from there it complains about the next\nrule in that format.\n\nIf you want to dig into this yourself I think the best way forward is\nthe last two paragraphs of\nhttps://public-inbox.org/git/20180824152016.20286-1-avarab@gmail.com/\n\n>> diff --git a/config.mak.uname b/config.mak.uname\n>> @@ -163,6 +163,10 @@ ifeq ($(uname_S),SunOS)\n>> BASIC_CFLAGS += -D__EXTENSIONS__ -D__sun__\n>> + # t/chainlint.sed is hopelessly broken all known (tested\n>> + # Solaris 10 & 11) versions of Solaris, both /bin/sed and\n>> + # /usr/xpg4/bin/sed\n>> + GIT_TEST_CHAIN_LINT = 0\n>> endif\n>>\n>> It slightly sucks to not have chainlint on\n>> Solaris, but it would also suck to revert chainlint.sed back to 2.18.0\n>> (there were some big improvements). So I think the patch above is the\n>> best way forward, especially since we're on rc2. What do you think?\n>\n> Keeping in mind that the main goal of 'chainlint' is to prevent _new_\n> breakage from entering the test suite, disabling 'chainlint' on\n> Solaris is an entirely reasonable way forward. In present day, it\n> seems quite unlikely that we'll see someone develop new tests on\n> Solaris, so having 'chainlint' disabled there isn't likely to be a big\n> deal. Moreover, if someone does write a new test on Solaris which has\n> a broken &&-chain in a subshell, that breakage will be caught very\n> quickly once the test is run by anyone on Linux or MacOS (or Windows\n> or BSD or AIX), so it's not like the broken test will make it into the\n> project undetected.\n\nSince writing my initial message I see that the CSW package packaging\ngit on Solaris just uses GNU userland:\nhttps://sourceforge.net/p/gar/code/HEAD/tree/csw/mgar/pkg/git/trunk/Makefile\n\nI.e. it puts /opt/csw/gnu in its $PATH, so this whole thing is probably\na non-issue. I.e. if the near-as-you-can-get \"official\" package just\ncompiles git against GNU tooling, perhaps we should stop caring about\nSolaris-native tooling.\n\nThat does also mean that something like my patch above isn't OK, since\nwe shouldn't be disabling GIT_TEST_CHAIN_LINT based on the uname, if it\nthen turns out that GNU tooling is being used.\n\nSo I think for the purposes of v2.19.0-rc2..v2.19.0 it's fine to just\nleave this as it is in master right now.\n\nIt's somewhat of a shame that we're drifting to not supporing some of\nthese more obscure POSIX systems \"natively\" though, but maybe that's the\nright move at this point, and maybe it isn't.\n\nWhen I was porting C code to Solaris ~10 years ago I'd often get SunCC\nturning up some interesting issues that were *real* issues, same with\ncompiling with IBM xlc on AIX.\n\nNow when I'm re-visiting these systems much later I've yet to turn up\nsome issue like that, just boring compatibility issues with their old\ntooling. E.g. one outstanding issue is that some test of ours fail on\nAIX because we use \"readlink\", but that command isn't in POSIX (only the\nC function is), so AIX doesn't implement it.\n\nSunCC used to be ahead of GCC & Clang when it came to certain classes of\nwarnings, but e.g. now everything it complains about is because it\ndoesn't understand C as well, e.g. we have quite a few compile warnings\ndue to code like this, which it claims is unreachable (but isn't):\nhttps://github.com/git/git/blob/v2.19.0-rc2/read-cache.c#L950-L955\n"},{"id":"357421","messageId":"nycvar.QRO.7.76.6.1809051536391.71@tvgsbejvaqbjf.bet","threadId":"49277","inReplyTo":"xmqqmusw6gbo.fsf@gitster-ct.c.googlers.com","subject":"jc/rebase-in-c-9-fixes, was Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Johannes Schindelin","fromEmail":"johannes.schindelin@gmx.de","sentAt":"2018-09-05T13:38:18Z","receivedAt":"2018-09-05T13:38:26Z","isPatch":false,"sender":{"key":"johannes.schindelin@gmx.de","avatar":"https://avatars.githubusercontent.com/u/127790?v=4"},"body":"Hi Junio,\n\nOn Tue, 4 Sep 2018, Junio C Hamano wrote:\n\n> * jc/rebase-in-c-9-fixes (2018-09-04) 1 commit\n> - rebase: re-add forgotten -k that stands for --keep-empty\n> (this branch uses ag/rebase-i-in-c, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final.)\n\nQuite frankly, I'd rather you pick up my two v2 series. I promised to take\nover from Pratik, who is busy with exams, and I did. I even announced as\nmuch in that mail where I pointed out the missing short option originally.\n\nSo there is no need to do add-on patches, as if the `rebase-in-c` patch\nseries had already been accepted into `next`.\n\nCiao,\nDscho\n"},{"id":"357442","messageId":"xmqq7ejz6gnn.fsf@gitster-ct.c.googlers.com","threadId":"49277","inReplyTo":"nycvar.QRO.7.76.6.1809051536391.71@tvgsbejvaqbjf.bet","subject":"Re: jc/rebase-in-c-9-fixes, was Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2018-09-05T16:41:16Z","receivedAt":"2018-09-05T16:41:20Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Johannes Schindelin writes:\n\n> On Tue, 4 Sep 2018, Junio C Hamano wrote:\n>\n>> * jc/rebase-in-c-9-fixes (2018-09-04) 1 commit\n>> - rebase: re-add forgotten -k that stands for --keep-empty\n>> (this branch uses ag/rebase-i-in-c, js/rebase-in-c-5.5-work-with-rebase-i-in-c, pk/rebase-in-c, pk/rebase-in-c-2-basic, pk/rebase-in-c-3-acts, pk/rebase-in-c-4-opts, pk/rebase-in-c-5-test and pk/rebase-in-c-6-final.)\n>\n> Quite frankly, I'd rather you pick up my two v2 series. I promised to take\n> over from Pratik, who is busy with exams, and I did. I even announced as\n> much in that mail where I pointed out the missing short option originally.\n\nYup, I'm planning to replace the large pile of patches with the new\nones you'd send once the final is done. The topic above is just a\nstop-gap measure to keep the tip of 'pu' testable, to be dropped\nwhen the series it fixes is replaced with a corrected one.\n"},{"id":"357443","messageId":"xmqqzhwv51wy.fsf@gitster-ct.c.googlers.com","threadId":"49277","inReplyTo":"CAPig+cRrwW662Hzp8V1sH51x8qYxPyqG=VHpOkfd-=RwUSSZbw@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2018-09-05T16:45:01Z","receivedAt":"2018-09-05T16:45:06Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Eric Sunshine writes:\n\n> On Wed, Sep 5, 2018 at 4:29 AM Ævar Arnfjörð Bjarmason wrote:\n>> I recently gained access to a Solaris 10 SPARC (5.10) box and discovered\n>> that the chainlint.sed implementation in 2.19.0 has more sed portability\n>> issues.\n>>\n>> First, whoever implemented the /bin/sed on Solaris apparently read the\n>> POSIX requirements for a max length label of 8 to mean that 8 characters\n>> should include the colon, so a bunch of things fail because of that, but\n>> are fixed with a shorter 7 character label.\n>\n> I'm pretty sure that Solaris 'sed' predates POSIX by a good bit, but\n> that's neither here nor there.\n\nI thought that we long time ago declared that it is a lost cause to\nuse Solaris with only tools in /bin and /usr/bin and instead depend\non /usr/xpg[46]/bin; did something happen recently to make us\nreconsider the position? If not, let's not waste time worrying\nabout /bin/sed over there.\n"},{"id":"357445","messageId":"CACsJy8CJfsjmt8Z03dT5enN86Ke7aMBHv+BvpRh8zFBgU4Z=Vw@mail.gmail.com","threadId":"49277","inReplyTo":"xmqqbm9b6gxs.fsf@gitster-ct.c.googlers.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Duy Nguyen","fromEmail":"pclouds@gmail.com","sentAt":"2018-09-05T16:48:06Z","receivedAt":"2018-09-05T16:48:35Z","isPatch":false,"sender":{"key":"pclouds@gmail.com","avatar":"https://avatars.githubusercontent.com/u/720?v=4"},"body":"On Wed, Sep 5, 2018 at 6:35 PM Junio C Hamano wrote:\n>\n> Junio C Hamano writes:\n>\n> > Here are the topics that have been cooking. Commits prefixed with\n> > '-' are only in 'pu' (proposed updates) while commits prefixed with\n> > '+' are in 'next'. The ones marked with '.' do not appear in any of\n> > the integration branches, but I am still holding onto them.\n> >\n> > Git 2.19-rc2 is out. Hopefully the tip of 'master' is more or less\n> > identical to the final one without needing much updates.\n>\n> By the way, linux-gcc job of TravisCI seems to have been unhappy\n> lately all the way down to 'master'. It fails split-index tests,\n> which may or may not be new breakage.\n>\n> https://travis-ci.org/git/git/jobs/424552273\n>\n> If this is a recent regression, we may want to revert a few commits,\n> but I do not offhand recall us having touched the spilt-index part\n> of the code during this cycle.\n\nI can't reproduce it here (with either 64 or 32 bit builds on x86).\nNot denying the problem, just a quick update. I'll need more time,\nmaybe over weekend to have a closer look.\n-- \nDuy\n"},{"id":"357458","messageId":"xmqqftyn4z4k.fsf@gitster-ct.c.googlers.com","threadId":"49277","inReplyTo":"CAPig+cTQLED6P2AWrxU=pvNc6WzO-5YVsPNh3fg28LdZqU15AA@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2018-09-05T17:45:15Z","receivedAt":"2018-09-05T17:45:20Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Eric Sunshine writes:\n\n> This description mischaracterizes what these changes are about.\n\nThanks for a replacement text; very much appreciated.\n\n"},{"id":"357462","messageId":"20180905181852.GA14181@localhost","threadId":"49277","inReplyTo":"CACsJy8CJfsjmt8Z03dT5enN86Ke7aMBHv+BvpRh8zFBgU4Z=Vw@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"SZEDER Gábor","fromEmail":"szeder.dev@gmail.com","sentAt":"2018-09-05T18:18:52Z","receivedAt":"2018-09-05T18:19:01Z","isPatch":false,"sender":{"key":"szeder.dev@gmail.com","avatar":"https://avatars.githubusercontent.com/u/116324?v=4"},"body":"On Wed, Sep 05, 2018 at 06:48:06PM +0200, Duy Nguyen wrote:\n> On Wed, Sep 5, 2018 at 6:35 PM Junio C Hamano wrote:\n> >\n> > Junio C Hamano writes:\n> >\n> > > Here are the topics that have been cooking. Commits prefixed with\n> > > '-' are only in 'pu' (proposed updates) while commits prefixed with\n> > > '+' are in 'next'. The ones marked with '.' do not appear in any of\n> > > the integration branches, but I am still holding onto them.\n> > >\n> > > Git 2.19-rc2 is out. Hopefully the tip of 'master' is more or less\n> > > identical to the final one without needing much updates.\n> >\n> > By the way, linux-gcc job of TravisCI seems to have been unhappy\n> > lately all the way down to 'master'. It fails split-index tests,\n> > which may or may not be new breakage.\n> >\n> > https://travis-ci.org/git/git/jobs/424552273\n> >\n> > If this is a recent regression, we may want to revert a few commits,\n> > but I do not offhand recall us having touched the spilt-index part\n> > of the code during this cycle.\n\nIt's not a regression, it's as old as the split index feature itself.\n\n> I can't reproduce it here (with either 64 or 32 bit builds on x86).\n> Not denying the problem, just a quick update. I'll need more time,\n> maybe over weekend to have a closer look.\n\nI have the patches almost ready, only the commit messages need some\ntouching up.\n\n"},{"id":"357464","messageId":"CAPig+cT-qZPRg2gjKU2o6TAzZDpdXm-mg6yk-LfsL+_a9jC5dw@mail.gmail.com","threadId":"49277","inReplyTo":"CAPig+cTQLED6P2AWrxU=pvNc6WzO-5YVsPNh3fg28LdZqU15AA@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Eric Sunshine","fromEmail":"sunshine@sunshineco.com","sentAt":"2018-09-05T18:44:08Z","receivedAt":"2018-09-05T18:44:23Z","isPatch":false,"sender":{"key":"sunshine@sunshineco.com","avatar":"https://avatars.githubusercontent.com/u/163641?v=4"},"body":"On Wed, Sep 5, 2018 at 5:14 AM Eric Sunshine wrote:\n> On Tue, Sep 4, 2018 at 6:36 PM Junio C Hamano wrote:\n> > * es/worktree-forced-ops-fix (2018-08-30) 9 commits\n> This description mischaracterizes what these changes are about. [...]\n>\n> So, perhaps rewrite this description like this:\n>\n> Fix a bug in which the same path could be registered under\n> multiple worktree entries if the patch was missing (for instance,\n> was removed manually).\n\nBah, typo: s/patch/path/\n\n> Also, as a convenience, expand the number of cases in which\n> --force is applicable.\n"},{"id":"357479","messageId":"20180905203111.GA9752@sigill.intra.peff.net","threadId":"49277","inReplyTo":"CAPig+cSETNc5nRxf_zQq6JE8zNyUvKmdPWNsm6=NMSOXm2DV=A@mail.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Jeff King","fromEmail":"peff@peff.net","sentAt":"2018-09-05T20:31:11Z","receivedAt":"2018-09-05T20:31:15Z","isPatch":false,"sender":{"key":"peff@peff.net","avatar":"https://avatars.githubusercontent.com/u/45925?v=4"},"body":"On Wed, Sep 05, 2018 at 05:50:04AM -0400, Eric Sunshine wrote:\n\n> On Tue, Sep 4, 2018 at 6:36 PM Junio C Hamano wrote:\n> > * es/worktree-forced-ops-fix (2018-08-30) 9 commits\n> > - worktree: delete .git/worktrees if empty after 'remove'\n> > - worktree: teach 'remove' to override lock when --force given twice\n> > - worktree: teach 'move' to override lock when --force given twice\n> > - worktree: teach 'add' to respect --force for registered but missing path\n> > - worktree: disallow adding same path multiple times\n> > - worktree: prepare for more checks of whether path can become worktree\n> > - worktree: generalize delete_git_dir() to reduce code duplication\n> > - worktree: move delete_git_dir() earlier in file for upcoming new callers\n> > - worktree: don't die() in library function find_worktree()\n> >\n> > Various subcommands of \"git worktree\" take '--force' but did not\n> > behave sensibly, which has been corrected.\n> \n> This series is missing a patch[1] from Peff which he wanted placed at\n> the end of the series. It was probably overlooked since he embedded\n> it as a reply in that thread rather than sending it as a standalone\n> patch.\n> \n> [1]: https://public-inbox.org/git/20180830075431.GF11944@sigill.intra.peff.net/\n\nYeah, I'm not sure which is easier for Junio. I figured by replying\ninline, it makes it easy to pick up on top of the others (since it\nreally does depend on them and should be in the same topic). But it is\neasier to overlook as \"meh, just more discussion\".\n\nEither way, yes, I'd be happy to see that patch on top.\n\n-Peff\n"},{"id":"357486","messageId":"xmqqlg8f38wl.fsf@gitster-ct.c.googlers.com","threadId":"49277","inReplyTo":"20180905203111.GA9752@sigill.intra.peff.net","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Junio C Hamano","fromEmail":"gitster@pobox.com","sentAt":"2018-09-05T21:56:58Z","receivedAt":"2018-09-05T21:57:03Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Jeff King writes:\n\n>> [1]: https://public-inbox.org/git/20180830075431.GF11944@sigill.intra.peff.net/\n>\n> Yeah, I'm not sure which is easier for Junio. I figured by replying\n> inline, it makes it easy to pick up on top of the others (since it\n> really does depend on them and should be in the same topic). But it is\n> easier to overlook as \"meh, just more discussion\".\n>\n> Either way, yes, I'd be happy to see that patch on top.\n\nThanks.\n"},{"id":"364011","messageId":"878t1i1d9q.fsf@evledraar.gmail.com","threadId":"49277","inReplyTo":"877ek0rymz.fsf@evledraar.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Ævar Arnfjörð Bjarmason","fromEmail":"avarab@gmail.com","sentAt":"2018-11-24T19:33:37Z","receivedAt":"2018-11-24T19:33:44Z","isPatch":false,"sender":{"key":"avarab@gmail.com","avatar":"https://avatars.githubusercontent.com/u/45301?v=4"},"body":"\nOn Wed, Sep 05 2018, Ævar Arnfjörð Bjarmason wrote:\n\n> On Wed, Sep 05 2018, Eric Sunshine wrote:\n>\n>> On Wed, Sep 5, 2018 at 4:29 AM Ævar Arnfjörð Bjarmason wrote:\n>>> I recently gained access to a Solaris 10 SPARC (5.10) box and discovered\n>>> that the chainlint.sed implementation in 2.19.0 has more sed portability\n>>> issues.\n>>>\n>>> First, whoever implemented the /bin/sed on Solaris apparently read the\n>>> POSIX requirements for a max length label of 8 to mean that 8 characters\n>>> should include the colon, so a bunch of things fail because of that, but\n>>> are fixed with a shorter 7 character label.\n>>\n>> I'm pretty sure that Solaris 'sed' predates POSIX by a good bit, but\n>> that's neither here nor there.\n>>\n>>> Then GIT_TEST_CHAIN_LINT=1 still fails because 878f988350 (\"t/test-lib:\n>>> teach --chain-lint to detect broken &&-chains in subshells\", 2018-07-11)\n>>> added a \"grep -q\" invocation. The /bin/grep on that version of Solaris\n>>> doesn't have -q.\n>>\n>> I knew that '-q' was potentially problematic on some platforms, so I\n>> checked and saw that existing tests were already using it, thus went\n>> ahead and used it. Dropping '-q' here and redirecting stderr to\n>> /dev/null is a perfectly fine alternative.\n>>\n>>> We fixed a similar issue way back in 80700fde91\n>>> (\"t/t1304: make a second colon optional in the mask ACL check\",\n>>> 2010-03-15) by redirecting to /dev/null instead.\n>>>\n>>> A bunch of other tests in the test suite rely on \"grep -q\", but nothing\n>>> as central as chainlint, so it makes everything break. Do we want to\n>>> away with \"grep -q\" entirely because of old Solaris /bin/grep?\n>>\n>> I count 132 instances in existing tests (though, I may have missed some).\n>>\n>>> At this point those familiar with Solaris are screaming (\"why are you\n>>> using anything in /bin!\"). Okey fine, but it mostly worked before, so\n>>> are we OK with breaking it? \"Mostly\" here is \"test suite would fail\n>>> 20-30 tests for various reasons, but at least no failures in test-lib.sh\n>>> and the like\".\n>>>\n>>> However, if as config.mak.uname does we run the tests with\n>>> PATH=/usr/xpg6/bin:/usr/xpg4/bin:$PATH, at this point sed is fine with 8\n>>> character labels [...]\n>>\n>> So, if you run the tests via 'make test' (or whatever), then you get\n>> /usr/xpg?/bin in PATH, but if you run an individual test script (and\n>> haven't set your PATH appropriately), then you encounter the problems\n>> you report above?\n>\n> You need to manually set the PATH before running the tests, the\n> SANE_TOOL_PATH just affects git-sh-setup. We should probably fix that,\n> i.e. now if you compile git our shellscripts will use the fixed paths,\n> but not the tests.\n>\n>>> [...] but starts complaining about this (also in\n>>> chainlint.sed):\n>>>\n>>> sed: Too many commands, last: s/\\n//\n>>\n>> Erm. Any additional context to help narrow this down?\n>\n> I tried playing around with it a bit, but honestly don't understand\n> enough of this advanced sed syntax to make any headway, it's complaining\n> about e.g. the \"check for multi-line double-quoted string\" rule, but\n> then if you remove the s/\\n// from there it complains about the next\n> rule in that format.\n>\n> If you want to dig into this yourself I think the best way forward is\n> the last two paragraphs of\n> https://public-inbox.org/git/20180824152016.20286-1-avarab@gmail.com/\n>\n>>> diff --git a/config.mak.uname b/config.mak.uname\n>>> @@ -163,6 +163,10 @@ ifeq ($(uname_S),SunOS)\n>>> BASIC_CFLAGS += -D__EXTENSIONS__ -D__sun__\n>>> + # t/chainlint.sed is hopelessly broken all known (tested\n>>> + # Solaris 10 & 11) versions of Solaris, both /bin/sed and\n>>> + # /usr/xpg4/bin/sed\n>>> + GIT_TEST_CHAIN_LINT = 0\n>>> endif\n>>>\n>>> It slightly sucks to not have chainlint on\n>>> Solaris, but it would also suck to revert chainlint.sed back to 2.18.0\n>>> (there were some big improvements). So I think the patch above is the\n>>> best way forward, especially since we're on rc2. What do you think?\n>>\n>> Keeping in mind that the main goal of 'chainlint' is to prevent _new_\n>> breakage from entering the test suite, disabling 'chainlint' on\n>> Solaris is an entirely reasonable way forward. In present day, it\n>> seems quite unlikely that we'll see someone develop new tests on\n>> Solaris, so having 'chainlint' disabled there isn't likely to be a big\n>> deal. Moreover, if someone does write a new test on Solaris which has\n>> a broken &&-chain in a subshell, that breakage will be caught very\n>> quickly once the test is run by anyone on Linux or MacOS (or Windows\n>> or BSD or AIX), so it's not like the broken test will make it into the\n>> project undetected.\n>\n> Since writing my initial message I see that the CSW package packaging\n> git on Solaris just uses GNU userland:\n> https://sourceforge.net/p/gar/code/HEAD/tree/csw/mgar/pkg/git/trunk/Makefile\n>\n> I.e. it puts /opt/csw/gnu in its $PATH, so this whole thing is probably\n> a non-issue. I.e. if the near-as-you-can-get \"official\" package just\n> compiles git against GNU tooling, perhaps we should stop caring about\n> Solaris-native tooling.\n>\n> That does also mean that something like my patch above isn't OK, since\n> we shouldn't be disabling GIT_TEST_CHAIN_LINT based on the uname, if it\n> then turns out that GNU tooling is being used.\n>\n> So I think for the purposes of v2.19.0-rc2..v2.19.0 it's fine to just\n> leave this as it is in master right now.\n>\n> It's somewhat of a shame that we're drifting to not supporing some of\n> these more obscure POSIX systems \"natively\" though, but maybe that's the\n> right move at this point, and maybe it isn't.\n>\n> When I was porting C code to Solaris ~10 years ago I'd often get SunCC\n> turning up some interesting issues that were *real* issues, same with\n> compiling with IBM xlc on AIX.\n>\n> Now when I'm re-visiting these systems much later I've yet to turn up\n> some issue like that, just boring compatibility issues with their old\n> tooling. E.g. one outstanding issue is that some test of ours fail on\n> AIX because we use \"readlink\", but that command isn't in POSIX (only the\n> C function is), so AIX doesn't implement it.\n>\n> SunCC used to be ahead of GCC & Clang when it came to certain classes of\n> warnings, but e.g. now everything it complains about is because it\n> doesn't understand C as well, e.g. we have quite a few compile warnings\n> due to code like this, which it claims is unreachable (but isn't):\n> https://github.com/git/git/blob/v2.19.0-rc2/read-cache.c#L950-L955\n\nCorrection. It is still useful for something:\nhttps://public-inbox.org/git/87a7ly1djb.fsf@evledraar.gmail.com/\n"},{"id":"364021","messageId":"20181125042835.GA24530@tor.lan","threadId":"49277","inReplyTo":"878t1i1d9q.fsf@evledraar.gmail.com","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Torsten Bögershausen","fromEmail":"tboegi@web.de","sentAt":"2018-11-25T04:28:35Z","receivedAt":"2018-11-25T04:28:44Z","isPatch":false,"sender":{"key":"tboegi@web.de","avatar":"https://avatars.githubusercontent.com/u/7138363?v=4"},"body":"On Sat, Nov 24, 2018 at 08:33:37PM +0100, Ævar Arnfjörð Bjarmason wrote:\n> \n> On Wed, Sep 05 2018, Ævar Arnfjörð Bjarmason wrote:\n> \n> > On Wed, Sep 05 2018, Eric Sunshine wrote:\n\n[]\n\n> > SunCC used to be ahead of GCC & Clang when it came to certain classes of\n> > warnings, but e.g. now everything it complains about is because it\n> > doesn't understand C as well, e.g. we have quite a few compile warnings\n> > due to code like this, which it claims is unreachable (but isn't):\n> > https://github.com/git/git/blob/v2.19.0-rc2/read-cache.c#L950-L955\n> \n\nWait a second - even if the compiler claims something (wrong)...\nthere a still 1+1/2 questions from my side:\n\n\nint verify_path(const char *path, unsigned mode)\n{\n\tchar c;\n\t ^\n\t/* Q1: should \"c\" be initialized like this: */\n\tchar c = *path;\n \n\tif (has_dos_drive_prefix(path))\n\t\treturn 0;\n\n\tgoto inside;\n\t^^^^^^^^^^^^ /* Q2: and why do we need the \"goto\" here ? */\n\tfor (;;) {\n\t\tif (!c)\n\t\t\treturn 1;\n\t\tif (is_dir_sep(c)) {\ninside:\n"},{"id":"364023","messageId":"20181125082145.GA15747@tor.lan","threadId":"49277","inReplyTo":"20181125042835.GA24530@tor.lan","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Torsten Bögershausen","fromEmail":"tboegi@web.de","sentAt":"2018-11-25T08:21:45Z","receivedAt":"2018-11-25T08:21:55Z","isPatch":false,"sender":{"key":"tboegi@web.de","avatar":"https://avatars.githubusercontent.com/u/7138363?v=4"},"body":"On Sun, Nov 25, 2018 at 05:28:35AM +0100, Torsten Bögershausen wrote:\n> On Sat, Nov 24, 2018 at 08:33:37PM +0100, Ævar Arnfjörð Bjarmason wrote:\n> > \n> > On Wed, Sep 05 2018, Ævar Arnfjörð Bjarmason wrote:\n> > \n> > > On Wed, Sep 05 2018, Eric Sunshine wrote:\n> \n> []\n> \n> > > SunCC used to be ahead of GCC & Clang when it came to certain classes of\n> > > warnings, but e.g. now everything it complains about is because it\n> > > doesn't understand C as well, e.g. we have quite a few compile warnings\n> > > due to code like this, which it claims is unreachable (but isn't):\n> > > https://github.com/git/git/blob/v2.19.0-rc2/read-cache.c#L950-L955\n> > \n> \n> Wait a second - even if the compiler claims something (wrong)...\n> there a still 1+1/2 questions from my side:\n> \n> \n> int verify_path(const char *path, unsigned mode)\n> {\n> \tchar c;\n> \t ^\n> \t/* Q1: should \"c\" be initialized like this: */\n> \tchar c = *path;\n> \n> \tif (has_dos_drive_prefix(path))\n> \t\treturn 0;\n> \n> \tgoto inside;\n> \t^^^^^^^^^^^^ /* Q2: and why do we need the \"goto\" here ? */\n> \tfor (;;) {\n> \t\tif (!c)\n> \t\t\treturn 1;\n> \t\tif (is_dir_sep(c)) {\n> inside:\n\nAfter some re-reading,\nI think that the \"goto inside\" was just hard to read....\n\nOut of interest:\nwould the following make the compiler happy ?\n\n\ndiff --git a/read-cache.c b/read-cache.c\nindex 49add63fe1..d574d58b9d 100644\n--- a/read-cache.c\n+++ b/read-cache.c\n@@ -951,17 +951,15 @@ static int verify_dotfile(const char *rest, unsigned mode)\n \n int verify_path(const char *path, unsigned mode)\n {\n-\tchar c;\n+\tchar c = *path ? '/' : '\\0';\n \n \tif (has_dos_drive_prefix(path))\n \t\treturn 0;\n \n-\tgoto inside;\n \tfor (;;) {\n \t\tif (!c)\n \t\t\treturn 1;\n \t\tif (is_dir_sep(c)) {\n-inside:\n \t\t\tif (protect_hfs) {\n \t\t\t\tif (is_hfs_dotgit(path))\n \t\t\t\t\treturn 0;\n"},{"id":"364035","messageId":"87zhtxz1ka.fsf@evledraar.gmail.com","threadId":"49277","inReplyTo":"20181125082145.GA15747@tor.lan","subject":"Re: What's cooking in git.git (Sep 2018, #01; Tue, 4)","fromName":"Ævar Arnfjörð Bjarmason","fromEmail":"avarab@gmail.com","sentAt":"2018-11-25T14:14:45Z","receivedAt":"2018-11-25T14:14:52Z","isPatch":false,"sender":{"key":"avarab@gmail.com","avatar":"https://avatars.githubusercontent.com/u/45301?v=4"},"body":"\nOn Sun, Nov 25 2018, Torsten Bögershausen wrote:\n\n> On Sun, Nov 25, 2018 at 05:28:35AM +0100, Torsten Bögershausen wrote:\n>> On Sat, Nov 24, 2018 at 08:33:37PM +0100, Ævar Arnfjörð Bjarmason wrote:\n>> >\n>> > On Wed, Sep 05 2018, Ævar Arnfjörð Bjarmason wrote:\n>> >\n>> > > On Wed, Sep 05 2018, Eric Sunshine wrote:\n>>\n>> []\n>>\n>> > > SunCC used to be ahead of GCC & Clang when it came to certain classes of\n>> > > warnings, but e.g. now everything it complains about is because it\n>> > > doesn't understand C as well, e.g. we have quite a few compile warnings\n>> > > due to code like this, which it claims is unreachable (but isn't):\n>> > > https://github.com/git/git/blob/v2.19.0-rc2/read-cache.c#L950-L955\n>> >\n>>\n>> Wait a second - even if the compiler claims something (wrong)...\n>> there a still 1+1/2 questions from my side:\n>>\n>>\n>> int verify_path(const char *path, unsigned mode)\n>> {\n>> \tchar c;\n>> \t ^\n>> \t/* Q1: should \"c\" be initialized like this: */\n>> \tchar c = *path;\n>>\n>> \tif (has_dos_drive_prefix(path))\n>> \t\treturn 0;\n>>\n>> \tgoto inside;\n>> \t^^^^^^^^^^^^ /* Q2: and why do we need the \"goto\" here ? */\n>> \tfor (;;) {\n>> \t\tif (!c)\n>> \t\t\treturn 1;\n>> \t\tif (is_dir_sep(c)) {\n>> inside:\n>\n> After some re-reading,\n> I think that the \"goto inside\" was just hard to read....\n>\n> Out of interest:\n> would the following make the compiler happy ?\n>\n>\n> diff --git a/read-cache.c b/read-cache.c\n> index 49add63fe1..d574d58b9d 100644\n> --- a/read-cache.c\n> +++ b/read-cache.c\n> @@ -951,17 +951,15 @@ static int verify_dotfile(const char *rest, unsigned mode)\n>\n> int verify_path(const char *path, unsigned mode)\n> {\n> -\tchar c;\n> +\tchar c = *path ? '/' : '\\0';\n>\n> \tif (has_dos_drive_prefix(path))\n> \t\treturn 0;\n>\n> -\tgoto inside;\n> \tfor (;;) {\n> \t\tif (!c)\n> \t\t\treturn 1;\n> \t\tif (is_dir_sep(c)) {\n> -inside:\n> \t\t\tif (protect_hfs) {\n> \t\t\t\tif (is_hfs_dotgit(path))\n> \t\t\t\t\treturn 0;\n\nI haven't tested (it's tedious) but yes, I can tell you SunCC won't\nwhine about this. I've only seen its unreachability detector get\nconfused about goto inside a loop, not the the sort of code you've\nreplaced this with.\n\nWe should not be appeasing these old compiler warnings in cases where\nthey're wrong. You can check out the CI output I linked to to see 10-20\ncases in the codebase where SunCC is wrong about unreliability.\n\nWhether we should just fix this for its own sake is another question.\n"}]}