Re: git-core: try_to_follow_renames(): git killed by SIGSEGV

[Konstantin Tokarev <annulen@yandex.ru>]

10.03.2020, 19:57, "SZEDER Gábor" <szeder.dev@gmail.com>:
> On Fri, Mar 06, 2020 at 03:44:34PM +0100, Alexandr Miloslavskiy wrote:
>>  Since I like studying crashes and noone else replied, I decided to have a
>>  look.
>>
>>  The problem is easy to reproduce with this (replace 1.c with any file):
>>    git log --follow -L 1,1:1.c -- 1.c
>
> Don't do this. In particular:
>
>   - Don't use line-level log with a pathspec, because the
>     documentation of 'git log -L' explicitly told you not to do so
>     ("You may not give any pathspec limiters."). This should have
>     errored out since the beginning, but, unfortunately, has never
>     been enforced.
>
>   - Don't use '-L' with '--follow'. On one hand, line-level log on
>     its own already follows file renames, even multiple files at once,
>     there is no need for an additional '--follow' (which can only
>     follow one file). OTOH, you shouldn't be able to use '-L' and
>     '--follow' together, because the former forbids a pathspec, while
>     the latter requires one.
>
> In any case, '--follow' has always been an ugly hack on top of the
> revision walking machinery, while line-level log is a rather poorly
> integrated bolt-on. They simply weren't designed to work together, as
> evidenced by their contradicting requirements about the pathspec.

This kind of explains issue with --follow and --full-diff which I've reported
recently and got completely ignored.

Are there any plans to integrate --follow better with other tools?


>
>>  It occurs because `opt->pathspec.items` gets cleaned here:
>>      clear_pathspec
>>      queue_diffs
>>          /* must look at the full tree diff to detect renames */
>>          clear_pathspec(&opt->pathspec);
>>          DIFF_QUEUE_CLEAR(&diff_queued_diff);
>>      process_ranges_ordinary_commit
>>      process_ranges_arbitrary_commit
>>      line_log_filter
>>      prepare_revision_walk
>>      cmd_log_walk
>>      cmd_log
>>
>>  And on next iteration it crashes in 'try_to_follow_renames' on this line:
>>      diff_opts.single_follow = opt->pathspec.items[0].match;
>>
>>  I think that bug comes from commit:
>>      a2bb801f by SZEDER Gábor, 2019-08-21 13:04:24
>>      line-log: avoid unnecessary full tree diffs
>>
>>  @szeder could you please look into that?
>>
>>  On 27.02.2020 13:56, Ondrej Pohorelsky wrote:
>>  >Hi,
>>  >
>>  >there is a SIGSEGV appearing in Fedora[0] with Git 2.24.1
>>  >
>>  >This bug started to appear after update to Git 2.24.1.
>>  >Bug reporter said that Git crashed on him while running VS Code with
>>  >Git Lens extension[1]
>>  >I have tried to reproduce this bug with my own compiled Git with debug
>>  >flags, but sadly SIGSEGV never appeared.
>>  >
>>  >To me it seems like there is a problem in commit a2bb801f6a[2] which
>>  >changes move_diff_queue() function. This function calls
>>  >diff_tree_oid() that calls try_to_follow_renames(). In the last two
>>  >functions there are no arguments checks.
>>  >
>>  >Best regards,
>>  >Ondřej Pohořelský
>>  >
>>  >[0] https://retrace.fedoraproject.org/faf/problems/bthash/?bth=25aa7d7267ab5de548ffca337115cb68f7b65105
>>  >[1] https://bugzilla.redhat.com/show_bug.cgi?id=1791810
>>  >[2] https://git.kernel.org/pub/scm/git/git.git/commit/?id=a2bb801f6a430f6049e5c9729a8f3bf9097d9b34
>>  >

-- 
Regards,
Konstantin