Re: Segfault in master due to 4fbcca4eff

On Fri, Sep 21 2018, Derrick Stolee wrote:

On 9/21/2018 10:30 AM, Ævar Arnfjörð Bjarmason wrote:

quoted

On Fri, Sep 21 2018, Junio C Hamano wrote:

quoted

* ds/reachable (2018-08-28) 19 commits
   (merged to 'next' on 2018-08-28 at b1634b371d)
  + commit-reach: correct accidental #include of C file
   (merged to 'next' on 2018-08-22 at 17f3275afb)
  + commit-reach: use can_all_from_reach
  + commit-reach: make can_all_from_reach... linear
  + commit-reach: replace ref_newer logic
  + test-reach: test commit_contains
  + test-reach: test can_all_from_reach_with_flags
  + test-reach: test reduce_heads
  + test-reach: test get_merge_bases_many
  + test-reach: test is_descendant_of
  + test-reach: test in_merge_bases
  + test-reach: create new test tool for ref_newer
  + commit-reach: move can_all_from_reach_with_flags
  + upload-pack: generalize commit date cutoff
  + upload-pack: refactor ok_to_give_up()
  + upload-pack: make reachable() more generic
  + commit-reach: move commit_contains from ref-filter
  + commit-reach: move ref_newer from remote.c
  + commit.h: remove method declarations
  + commit-reach: move walk methods from commit.c

  The code for computing history reachability has been shuffled,
  obtained a bunch of new tests to cover them, and then being
  improved.

There's a segfault now in master when fetching because of 4fbcca4eff
("commit-reach: make can_all_from_reach... linear", 2018-07-20). I have
not had time to debug this, or found an easy isolated test case, but
this script will reliably make it segfault for me:

     #!/bin/sh

     git_dot_git=/home/avar/g/git

     old=v0.99
     new=v0.99.1

     rm -rf /tmp/srv
     rm -rf /tmp/cln
     git init --bare /tmp/srv
     git init --bare /tmp/cln
     $git_dot_git/git --exec-path=$git_dot_git push file:///tmp/srv $old:refs/whatever/ref
     $git_dot_git/git --exec-path=$git_dot_git -C /tmp/cln fetch file:///tmp/srv 'refs/*:refs/*'
     $git_dot_git/git --exec-path=$git_dot_git push file:///tmp/srv $new:refs/whatever/ref
     if GIT_TRACE=1 $git_dot_git/git --exec-path=$git_dot_git -C /tmp/cln fetch file:///tmp/srv 'refs/*:refs/*'
     then
         exit 0
     else
         exit 1
     fi

I.e. you need to push the v0.99 tag to its own repo, fetch that from
another one, then push v0.99.1, fetch everything, and you'll get a
segfault:

     remote: Resolving deltas: 100% (187/187), completed with 48 local objects.
     To file:///tmp/srv
        d6602ec519..f25a265a34  v0.99.1 -> refs/whatever/ref
     14:26:44.505787 git.c:415               trace: built-in: git fetch file:///tmp/srv 'refs/*:refs/*'
     14:26:44.506708 run-command.c:637       trace: run_command: unset GIT_DIR GIT_IMPLICIT_WORK_TREE GIT_PREFIX; 'git-upload-pack '\''/tmp/srv'\'''
     14:26:44.508831 git.c:415               trace: built-in: git upload-pack /tmp/srv
     14:26:44.509953 run-command.c:637       trace: run_command: git rev-list --objects --stdin --not --all --quiet
     Segmentation fault
     fatal: The remote end hung up unexpectedly

Same with refs/tags/ref b.t.w., not just refs/whatever/ref, I just was
initially testing this for some follow-up work on my series for checking
how fetching to various non-heads/tags namespaces works.

This error was reported by Peff [1] and fixed in [2], but as stated
[3] I was waiting for more review before sending a v3. I'll send that
v3 shortly, responding to the feedback so far.

-Stolee

[1]
https://public-inbox.org/git/0bf9103c-9377-506b-7ad7-e5273d8e94fc@gmail.com/T/#u

[2] https://public-inbox.org/git/pull.39.git.gitgitgadget@gmail.com/

[3]
https://public-inbox.org/git/8d6061de-1654-577c-40c6-211dbd03aa36@gmail.com/

Thanks and sorry for the duplicate report. I can confirm that applying
the v2 of that fixes the segfault for the test case I posted.