regression - fails to drop . Exit code 11 wo --debug, and 1 with --debug

Please describe the problem.

DataLad unittest (datalad/crawler/pipelines/tests/test_simple_s3.py:test_drop) checking communication of git-annex with our special remote started to hang (should have crashed I guess, yet to check) and the attempt in independent execution showed that git-annex simply crashes. Downgrade from 6.20180506+gitg6316072f0-1~ndall+1 to 6.20180416+gitg86b18966f-1~ndall+1 resolved the issue. Seems to boil down to change on interactions with the remote with new version crashing with git-annex: readDirStream: invalid argument (Bad file descriptor)?

May be manages to stop/kill the custom special remote before stopping communicating with it?

Please provide any additional information below.

$> git annex whereis --all --debug
[2018-05-07 21:49:18.991105877] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","git-annex"]
[2018-05-07 21:49:18.997322576] process done ExitSuccess
[2018-05-07 21:49:18.997426863] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","--hash","refs/heads/git-annex"]
[2018-05-07 21:49:19.003137341] process done ExitSuccess
[2018-05-07 21:49:19.003344009] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","log","refs/heads/git-annex..f9909eda54827d61b254605b5631482b3dc65468","--pretty=%H","-n1"]
[2018-05-07 21:49:19.007972053] process done ExitSuccess
[2018-05-07 21:49:19.008717245] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch"]
[2018-05-07 21:49:19.009717171] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch-check=%(objectname) %(objecttype) %(objectsize)"]
whereis MD5E-s8--966634ebf2fc135707d6753692bf4b1e.txt [2018-05-07 21:49:19.016115806] chat: /home/yoh/proj/datalad/datalad-master/venvs/dev/bin/git-annex-remote-datalad []
[2018-05-07 21:49:19.832279118] git-annex-remote-datalad[1] --> VERSION 1
[2018-05-07 21:49:19.832466638] git-annex-remote-datalad[1] <-- EXTENSIONS INFO
[2018-05-07 21:49:19.832654467] git-annex-remote-datalad[1] --> UNSUPPORTED-REQUEST
[2018-05-07 21:49:19.832898699] git-annex-remote-datalad[1] <-- PREPARE
[2018-05-07 21:49:19.83308154] git-annex-remote-datalad[1] --> PREPARE-SUCCESS
[2018-05-07 21:49:19.833153568] git-annex-remote-datalad[1] <-- WHEREIS MD5E-s8--966634ebf2fc135707d6753692bf4b1e.txt
[2018-05-07 21:49:19.833276243] git-annex-remote-datalad[1] --> WHEREIS-FAILURE
[2018-05-07 21:49:19.833536457] git-annex-remote-datalad[1] <-- CLAIMURL s3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null
[2018-05-07 21:49:19.833847975] git-annex-remote-datalad[1] --> DEBUG Claiming url 's3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null'
[2018-05-07 21:49:19.833975294] Claiming url 's3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null'
[2018-05-07 21:49:19.834045759] git-annex-remote-datalad[1] --> CLAIMURL-SUCCESS
(1 copy) 
    cf13d535-b47c-5df6-8590-0793cb08a90a -- [datalad]

  datalad: s3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null
ok
git-annex: readDirStream: invalid argument (Bad file descriptor)

(dev) 1 19899 ->1.....................................:Mon 07 May 2018 09:49:19 PM EDT:.

# so -- it exited with exit code 1.  Downgrading now!

$> sudo apt-get install git-annex-standalone=6.20180416+gitg86b18966f-1~ndall+1        
...

$> git annex whereis --all --debug                                                     
[2018-05-07 21:50:33.064546504] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","git-annex"]
[2018-05-07 21:50:33.075890599] process done ExitSuccess
[2018-05-07 21:50:33.076053279] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","--hash","refs/heads/git-annex"]
[2018-05-07 21:50:33.082610193] process done ExitSuccess
[2018-05-07 21:50:33.082774661] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","log","refs/heads/git-annex..f9909eda54827d61b254605b5631482b3dc65468","--pretty=%H","-n1"]
[2018-05-07 21:50:33.088684568] process done ExitSuccess
[2018-05-07 21:50:33.089343824] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch"]
[2018-05-07 21:50:33.090037985] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch-check=%(objectname) %(objecttype) %(objectsize)"]
[2018-05-07 21:50:33.09515616] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","ls-tree","--full-tree","-z","-r","--name-only","--","refs/heads/git-annex"]
whereis MD5E-s8--966634ebf2fc135707d6753692bf4b1e.txt [2018-05-07 21:50:33.101385386] chat: /home/yoh/proj/datalad/datalad-master/venvs/dev/bin/git-annex-remote-datalad []
[2018-05-07 21:50:33.935964594] git-annex-remote-datalad[1] --> VERSION 1
[2018-05-07 21:50:33.936100411] git-annex-remote-datalad[1] <-- EXTENSIONS INFO
[2018-05-07 21:50:33.936325748] git-annex-remote-datalad[1] --> UNSUPPORTED-REQUEST
[2018-05-07 21:50:33.936398107] git-annex-remote-datalad[1] <-- PREPARE
[2018-05-07 21:50:33.936521005] git-annex-remote-datalad[1] --> PREPARE-SUCCESS
[2018-05-07 21:50:33.936577067] git-annex-remote-datalad[1] <-- WHEREIS MD5E-s8--966634ebf2fc135707d6753692bf4b1e.txt
[2018-05-07 21:50:33.936676987] git-annex-remote-datalad[1] --> WHEREIS-FAILURE
[2018-05-07 21:50:33.937304604] git-annex-remote-datalad[1] <-- CLAIMURL s3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null
[2018-05-07 21:50:33.937515713] git-annex-remote-datalad[1] --> DEBUG Claiming url 's3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null'
[2018-05-07 21:50:33.937580054] Claiming url 's3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null'
[2018-05-07 21:50:33.937645676] git-annex-remote-datalad[1] --> CLAIMURL-SUCCESS
(1 copy) 
    cf13d535-b47c-5df6-8590-0793cb08a90a -- [datalad]

  datalad: s3://datalad-test0-nonversioned/1version-nonversioned1.txt?versionId=null
ok
[2018-05-07 21:50:33.973870482] process done ExitSuccess
[2018-05-07 21:50:33.974304561] process done ExitSuccess
[2018-05-07 21:50:33.97461093] process done ExitSuccess

all kosher.

Original manifestation was (note that exit code was 1 for with --debug and 139 (128+11) without --debug):

(git-annex)hopa:~/.tmp/datalad_temp_test_drop2bWvne[master]git
$> git annex drop --debug --json --all --force --
[2018-05-07 21:43:12.673253137] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","git-annex"]
[2018-05-07 21:43:12.677379888] process done ExitSuccess
[2018-05-07 21:43:12.677452681] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","show-ref","--hash","refs/heads/git-annex"]
[2018-05-07 21:43:12.681153081] process done ExitSuccess
[2018-05-07 21:43:12.681279588] read: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","log","refs/heads/git-annex..f9909eda54827d61b254605b5631482b3dc65468","--pretty=%H","-n1"]
[2018-05-07 21:43:12.686123605] process done ExitSuccess
[2018-05-07 21:43:12.686850715] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch"]
[2018-05-07 21:43:12.687525693] chat: git ["--git-dir=.git","--work-tree=.","--literal-pathspecs","cat-file","--batch-check=%(objectname) %(objecttype) %(objectsize)"]
git-annex: readDirStream: invalid argument (Bad file descriptor)
(dev) 1 19888 ->1.....................................:Mon 07 May 2018 09:43:12 PM EDT:.
(git-annex)hopa:~/.tmp/datalad_temp_test_drop2bWvne[master]git
$> git annex drop  --json --all --force --     
error: git-annex died of signal 11
(dev) 1 19889 ->139.....................................:Mon 07 May 2018 09:43:16 PM EDT:.

which also resolves with downgrade of git-annex.

fixed --Joey

RSS Atom

comment 1

Neither git rev 6316072f0 nor 86b18966f is present in my git-annex git repo, which makes it kind of hard to look at the diff between whatever versions of git-annex you're running. Are you building out of a patched git repository?

My guess is bea0ad220a68138dc0a43d0c86bb2353ecf92d2c since it involves both directory reading and unsafeInterleaveIO.

Indeed, it looks like that could defer a readDirStream due to unsafeInterleaveIO to outside the open/close bracket. And I suppose reading from a closed file handle might conceivably segfault.

Looks like that would affect bare repositories, or using --all. I don't think it would affect those commands in a non-bare repository. Ah, I see youre whereis is indeed in a bare repository, and your drop used --all, so my analysis seems right.

Fixed by backing out the problematic part of the commit.

Comment by joey — Tue May 8 17:09:35 2018

Remove comment