# [BUG] Problem with "stgit push" causing data loss 4 messages from 2007-01-20 to 2007-01-21. Participants: Yann Dirson, Catalin Marinas. Thread: https://gitlist.dev/t/6453 ## Yann Dirson, 2007-01-20 15:01 Subject: [BUG] Problem with "stgit push" causing data loss Message-ID: <20070120150113.GB4665@nan92-1-81-57-214-146.fbx.proxad.net> URL: https://gitlist.dev/e/20070120150113.GB4665%40nan92-1-81-57-214-146.fbx.proxad.net ``` Well, "data loss" is a bit strong, since the data is still available to find using git-lost-found. I first tried to reproduce with a simple test-case, but could not, so here is my way to reproduce it, on a clone of the stgit official workspace. stgit$ stg branch --create test origin Branch "test" created. The test patch is my previous work on "pull --to", now superceded by "rebase". Patch to reproduce attached to this mail - I just use "pick" since it's easier for me: stgit$ ./stg pick -n pull-to 9abaec589753ab190d378b534ceaa6a5af5d0dd3 Checking for changes in the working directory... done Importing commit 9abaec589753ab190d378b534ceaa6a5af5d0dd3... done Now at patch "pull-to" stgit$ stg files pull-to M stgit/commands/pull.py A t/t2100-pull-to.sh Now move the testcase in a new patch below, using my stg-fold-files-from already sent to the list: stgit$ stg pop Popping patch "pull-to"... done No patches applied stgit$ stg new test -m test stgit$ stg-fold-files-from pull-to 't/*' Folding patch from stdin... done stgit$ stg refresh Refreshing patch "test"... done stgit$ stg files A t/t2100-pull-to.sh And then push the old file over that: $ stg push Pushing patch "pull-to"... Traceback (most recent call last): File "/usr/bin/stg", line 43, in ? main() File "/var/lib/python-support/python2.4/stgit/main.py", line 261, in main command.func(parser, options, args) File "/var/lib/python-support/python2.4/stgit/commands/push.py", line 101, in func push_patches(patches, options.merged) File "/var/lib/python-support/python2.4/stgit/commands/common.py", line 190, in push_patches modified = crt_series.push_patch(p) File "/var/lib/python-support/python2.4/stgit/stack.py", line 916, in push_patch git.merge(bottom, head, top) File "/var/lib/python-support/python2.4/stgit/git.py", line 535, in merge stages['2'][0], stages['3'][0]) != 0: File "/var/lib/python-support/python2.4/stgit/gitmergeonefile.py", line 203, in merge print >> sys.stderr, \ TypeError: not all arguments converted during string formatting State of things now: stgit$ stg status ? t/t2100-pull-to.sh.current ? t/t2100-pull-to.sh.patched M stgit/commands/pull.py stgit$ diff t/t2100-pull-to.sh.*; echo $? 0 stgit$ stg files pull-to stgit$ stg top test As you can see, the pull-to patch is now in a situation that's not obvious to untangle by hand. Also "stg push -m" exhibits the same issue. Best regards, -- Yann. ``` ## Yann Dirson, 2007-01-20 15:09 Subject: Re: [BUG] Problem with "stgit push" causing data loss Message-ID: <20070120150949.GB4684@nan92-1-81-57-214-146.fbx.proxad.net> URL: https://gitlist.dev/e/20070120150949.GB4684%40nan92-1-81-57-214-146.fbx.proxad.net In-Reply-To: <20070120150113.GB4665@nan92-1-81-57-214-146.fbx.proxad.net> ``` On Sat, Jan 20, 2007 at 04:01:13PM +0100, Yann Dirson wrote: > The test patch is my previous work on "pull --to", now superceded > by "rebase". Patch to reproduce attached to this mail Well... to *this* mail indeed :} commit 9abaec589753ab190d378b534ceaa6a5af5d0dd3 Author: Yann Dirson Date: Wed Jan 17 22:11:58 2007 +0100 Pull to arbitrary commit diff --git a/stgit/commands/pull.py b/stgit/commands/pull.py index 7c5db22..865bfd6 100644 --- a/stgit/commands/pull.py +++ b/stgit/commands/pull.py @@ -42,7 +42,9 @@ options = [make_option('-n', '--nopush', action = 'store_true'), make_option('-m', '--merged', help = 'check for patches merged upstream', - action = 'store_true')] + action = 'store_true'), + make_option('--to', metavar = 'COMMITID', + help = 'move the stack base to COMMITID')] def func(parser, options, args): """Pull the changes from a remote repository @@ -78,10 +80,15 @@ def func(parser, options, args): crt_series.pop_patch(applied[0]) print 'done' - # pull the remote changes - print 'Pulling from "%s"...' % repository - git.pull(repository, refspec) - print 'done' + # FIXME: check git_id/rev_parse(options.to) first and maybe try to fetch if not found + if options.to: + print 'Rebasing to "%s"...' % options.to + git.move_branch(options.to) + else: + # pull the remote changes + print 'Pulling from "%s"...' % repository + git.pull(repository, refspec) + print 'done' # push the patches back if not options.nopush: diff --git a/t/t2100-pull-to.sh b/t/t2100-pull-to.sh new file mode 100755 index 0000000..0bf5f16 --- /dev/null +++ b/t/t2100-pull-to.sh @@ -0,0 +1,58 @@ +#!/bin/sh +# +# Copyright (c) 2007 Yann Dirson +# + +test_description='Test the "pull --to" command.' + +. ./test-lib.sh + +# don't need this repo, but better not drop it, see t1100 +#rm -rf .git + +# Need a repo to clone +test_create_repo foo + +test_expect_success \ + 'Setup and clone tree, and setup changes' \ + 'cd foo && + stg init && stg new patch -m . && + printf "a\nb\n" > file && stg add file && stg refresh && + cd .. && + stg clone foo bar && + cd bar && + stg new p1 -m p1 && + echo "c" > file2 && stg add file2 && stg refresh && + cd .. && + cp -a bar bar2 +' + +test_expect_success \ + 'Modify original stack' \ + '(cd foo && echo "c" >> file && stg refresh)' + +#test_debug bash +#test_done + +#test_debug "cd foo && gitk --all && cd .." + +test_expect_success \ + 'Pull without --to' \ + '(cd bar && + a=$(git rev-parse HEAD) && + stg pull && + test "$a" == $(git rev-parse HEAD))' + +#test_debug "cd bar && gitk --all && cd .." + +test_expect_success \ + 'Pull with --to' \ + '(cd bar2 && + a=$(git rev-parse HEAD) && + git fetch origin +refs/heads/master:refs/heads/origin && + stg pull --to origin && + test "$a" != $(git rev-parse HEAD) )' + +test_debug "cd bar2 && gitk --all && cd .." + +test_done ``` ## Catalin Marinas, 2007-01-21 22:42 Subject: Re: [BUG] Problem with "stgit push" causing data loss Message-ID: URL: https://gitlist.dev/e/b0943d9e0701211442q503396ack2bf2d088325fc4a0%40mail.gmail.com In-Reply-To: <20070120150113.GB4665@nan92-1-81-57-214-146.fbx.proxad.net> ``` On 20/01/07, Yann Dirson wrote: > Well, "data loss" is a bit strong, since the data is still available > to find using git-lost-found. > > I first tried to reproduce with a simple test-case, but could not, so > here is my way to reproduce it, on a clone of the stgit official > workspace. I tried to somehow emulate this without the fold script but didn't manage to trigger the bug. I'll try with your script tomorrow but first: Do you use the latest HEAD in the StGIT repository? I'm asking because the line numbers reported in your backtrace are different from mine. I also made some changes in this area in the past couple of weeks which might have fixed or broken things. -- Catalin ``` ## Yann Dirson, 2007-01-21 23:27 Subject: Re: [BUG] Problem with "stgit push" causing data loss Message-ID: <20070121232745.GE4665@nan92-1-81-57-214-146.fbx.proxad.net> URL: https://gitlist.dev/e/20070121232745.GE4665%40nan92-1-81-57-214-146.fbx.proxad.net In-Reply-To: ``` On Sun, Jan 21, 2007 at 10:42:18PM +0000, Catalin Marinas wrote: > On 20/01/07, Yann Dirson wrote: > >Well, "data loss" is a bit strong, since the data is still available > >to find using git-lost-found. > > > >I first tried to reproduce with a simple test-case, but could not, so > >here is my way to reproduce it, on a clone of the stgit official > >workspace. > > I tried to somehow emulate this without the fold script but didn't > manage to trigger the bug. I'll try with your script tomorrow but > first: > > Do you use the latest HEAD in the StGIT repository? I'm asking because > the line numbers reported in your backtrace are different from mine. I > also made some changes in this area in the past couple of weeks which > might have fixed or broken things. Here is where I am: stgit$ git-rev-parse origin 0f4eba6a37c1a5454560b097873e5a22bfcde908 ```