# What's in git.git (stable) 3 messages from 2007-08-11 to 2007-08-16. Participants: Junio C Hamano, David Kastrup. Thread: https://gitlist.dev/t/9481 ## Junio C Hamano, 2007-08-11 08:41 Subject: What's in git.git (stable) Message-ID: <7vir7mo42w.fsf@assigned-by-dhcp.cox.net> URL: https://gitlist.dev/e/7vir7mo42w.fsf%40assigned-by-dhcp.cox.net ``` I thought we were pretty in good shape and in a nice and quiet freeze period. Until a few days ago. Then suddenly, flurry of activity happened. A few performance issues were raised and fixed: * "git-status" on a huge tree was way suboptimal and found to have unnecessary O(n^2) codepath. Fixing this also sped up "git-diff --cached"; * "git-commit paths..." had a few other bottlenecks. "git-add --stdin" was one of them; * In addition, Linus optimized all three cases of "git read-tree" that had the same inefficiency; * "git-bundle create" had a stupid "one-byte-at-a-time" loop that was unnecessary. Also one of the new features in 1.5.3, GIT_WORK_TREE, was found to be not-quite-ready. I think the few commits during the last couple of days should finally make it ready. I have resisted merging any new features that were not present in 1.5.3-rc1, but it appears that we would need to have a few more -rc rounds before the final _anyway_, so I went into a merge-binge. A couple of well done topics from the next branch are now in master: * Carlos's "builtin tag" series; * David's "format documentation in info format as well" series; Expect the tip of 'master' to be tagged v1.5.3-rc5 and let's hope it to be the last -rc before 1.5.3 final. ---------------------------------------------------------------- * The 'master' branch has these since the last announcement in addition to the above. Alex Riesen (1): Fix filehandle leak in "git branch -D" Brian Downing (1): cvsserver: Fix for work trees Carlos Rica (3): Make git tag a builtin. builtin-tag.c: Fix two memory leaks and minor notation changes. Make verify-tag a builtin. David Kastrup (4): Documentation/git-commit.txt: correct bad list formatting. Add support for an info version of the user manual INSTALL: explain info installation and dependencies. Documentation/Makefile: remove cmd-list.made before redirecting to it. Johannes Schindelin (3): launch_editor(): Heed GIT_EDITOR and core.editor settings Teach "git stripspace" the --strip-comments option Reinstate the old behaviour when GIT_DIR is set and GIT_WORK_TREE is unset Junio C Hamano (9): git-clone: aggressively optimize local clone behaviour. Reorder the list of commands in the manual. Fix formatting of git-blame documentation. Fix an illustration in git-rev-parse.txt tweak manpage formatting Revert "tweak manpage formatting" Optimize "diff --cached" performance. allow git-bundle to create bottomless bundle allow git-bundle to create bottomless bundle Linus Torvalds (7): connect: accept file:// URL scheme Start moving unpack-trees to "struct tree_desc" Fix "git commit directory/" performance anomaly Move old index entry removal from "unpack_trees()" into the individual functions Optimize the common cases of git-read-tree Optimize the two-way merge of git-read-tree too Optimize the three-way merge of git-read-tree Mark Levedahl (2): builtin-bundle.c - use stream buffered input for rev-list builtin-bundle - use buffered reads for bundle header Shawn O. Pearce (3): Teach update-paranoid how to store ACLs organized by groups Teach the update-paranoid to look at file differences Use the empty tree for base diff in paranoid-update on new branches Simon Hausmann (2): git-p4: Fix support for symlinks. git-p4: Fix git-p4 submit to include only changed files in the perforce submit template. Steve Hoelzer (2): git-stash documentation: stash numbering starts at zero, not one git-stash documentation: add missing backtick Steven Grimm (1): Add a note about the index being updated by git-status in some cases Uwe Kleine-König (2): send-email: rfc822 forbids using without a non-empty "phrase" send-email: get all the quoting of realnames right ``` ## David Kastrup, 2007-08-11 09:32 Subject: Re: What's in git.git (stable) Message-ID: <85hcn6xvqm.fsf@lola.goethe.zz> URL: https://gitlist.dev/e/85hcn6xvqm.fsf%40lola.goethe.zz In-Reply-To: <7vir7mo42w.fsf@assigned-by-dhcp.cox.net> ``` Junio C Hamano writes: > I thought we were pretty in good shape and in a nice and quiet > freeze period. > > Until a few days ago. > > Then suddenly, flurry of activity happened. A few performance > issues were raised and fixed: The fix to git-filter-branch I posted is tested and obviously necessary (without it, git-filter-branch will only work if GIT_DIR is set previously to calling it). Also documentation and usage of git-filter-branch are out of kilter, the given examples for wiping a file from history don't work, specifing the refs to work on reacts impredictably (it is not clear what gets accepted and what not, and one seemingly has to always specify "HEAD" which is basically ignored). Could you ask Dscho to consider adding cases based on the documentation to the test suite? That should help weed out the worst discrepances. -- David Kastrup, Kriemhildstr. 15, 44793 Bochum ``` ## Junio C Hamano, 2007-08-16 05:02 Subject: What's in git.git (stable) Message-ID: <7v1we4m5q8.fsf@gitster.siamese.dyndns.org> URL: https://gitlist.dev/e/7v1we4m5q8.fsf%40gitster.siamese.dyndns.org In-Reply-To: <7vir7mo42w.fsf@assigned-by-dhcp.cox.net> ``` Maintenance release 1.5.2.5 is out, so that a fix for "add -u" breakage can be propagated out to a released version without waiting for the 1.5.3 final. The 'master' branch has a few more fixes since -rc5. Notable are: - fix "add -u" (in 'maint'); - fix "read-tree -m base1 base2 ours theirs"; - fix "clone --bare"; - teach "git apply" to handle submodule patches; I would like to ask subsystem people to give a pull request destined for the 1.5.3 final by the end of the week, so that we can do -rc6 and make it the last rc in this cycle. ---------------------------------------------------------------- * The 'maint' branch has these fixes since the last announcement. Junio C Hamano (2): Fix "git add -u" data corruption. GIT 1.5.2.5 Salikh Zakirov (1): git-add -u paths... now works from subdirectory * The 'master' branch has these since the last announcement in addition to the above. Alberto Bertogli (1): Allow git-svnimport to take "" as the trunk directory. Alex Riesen (2): gitk: Continue and show error message in new repos gitk: Show an error and exit if no .git could be found Alexandre Julliard (3): git.el: Add support for interactive diffs. git.el: Always set the current directory in the git-diff buffer. git-add: Add support for --refresh option. Brian Downing (1): Add read_cache to builtin-check-attr Brian Gernhardt (1): Fix t5701-clone-local for white space from wc David Kastrup (2): git-sh-setup.sh: make GIT_DIR absolute Add a test for git-commit being confused by relative GIT_DIR Eric Wong (1): git-svn: fix log with single revision against a non-HEAD branch Junio C Hamano (10): t3404: fix "fake-editor" builtin-bundle create - use lock_file git-diff: squelch "empty" diffs merge-recursive: do not rudely die on binary merge attr.c: refactoring attr.c: read .gitattributes from index as well. GIT 1.5.3-rc5 Fix read-tree merging more than 3 trees using 3-way merge Update documentation links for older releases. git-clone: allow --bare clone Luiz Fernando N. Capitulino (3): Avoid ambiguous error message if pack.idx header is wrong Introduces xmkstemp() Use xmkstemp() instead of mkstemp() Marco Costalba (1): Add --log-size to git log to print message size Mark Levedahl (3): gitk: Enable selected patch text on Windows gitk: Handle MouseWheel events on Windows t3902 - skip test if file system doesn't support HT in names Nicolas Pitre (1): pack-objects: remove bogus arguments to delta_cacheable() Paul Mackerras (4): gitk: Add a context menu for file list entries gitk: Fix bug causing the "can't unset idinlist(...)" error gitk: Fix bug introduced in commit 67a4f1a7 gitk: Fix bug causing Tcl error when updating graph Reece H. Dunn (1): git-p4: Fix the sorting of changelists when cloning a Perforce repository. René Scharfe (3): diff: don't run pager if user asked for a diff style exit code diff: squelch empty diffs even more path-list.c: always free strdup'ed paths Steffen Prohaska (1): Improved hint on how to set identity Sven Verdoolaege (1): git-apply: apply submodule changes ```