problems GIT_TRACE

problems GIT_TRACE

[Matthias Lederhofer]

I just discovered a problem with GIT_TRACE.  Some scripts redirect
stderr to stdout and the trace messages go with it.  For example from
git-repack:
> name=$(git-rev-list --objects --all $rev_list 2>&1 |
>         git-pack-objects --non-empty $pack_objects .tmp-pack) ||
>         exit 1
Then for example git-pack-objects complains:
> fatal: expected sha1, got garbage:
>  trace: built-in: git 'rev-list' '--objects' '--all'
git-grep '2>&' shows a few other places that do this too, I'll take a
closer look at this later.  Is there any reason to redirect stderr to
stdout?  I think this will always fail with such a strange error
message when something is written to stderr.

Re: problems GIT_TRACE

[Junio C Hamano]

Matthias Lederhofer <matled@gmx.net> writes:

> I just discovered a problem with GIT_TRACE.  Some scripts redirect
> stderr to stdout and the trace messages go with it.  For example from
> git-repack:
>> name=$(git-rev-list --objects --all $rev_list 2>&1 |
>>         git-pack-objects --non-empty $pack_objects .tmp-pack) ||
>>         exit 1
> Then for example git-pack-objects complains:
>> fatal: expected sha1, got garbage:
>>  trace: built-in: git 'rev-list' '--objects' '--all'
> git-grep '2>&' shows a few other places that do this too, I'll take a
> closer look at this later.  Is there any reason to redirect stderr to
> stdout?  I think this will always fail with such a strange error
> message when something is written to stderr.

This particular one is trying to catch an error condition from
rev-list.  Shell reports the exit status from the last command
in the pipeline, and when rev-list notices a corrupt repository,
it wants to tell pack-objects to stop producing incorrect pack,
but there is no way other than sending a garbage string (as an
error message) to cause pack-object to notice there is garbage
coming in.

[PATCH] fix git-repack for use with GIT_TRACE

[Matthias Lederhofer]

---
Junio C Hamano <junkio@cox.net> wrote:
> This particular one is trying to catch an error condition from
> rev-list.  Shell reports the exit status from the last command
> in the pipeline, and when rev-list notices a corrupt repository,
> it wants to tell pack-objects to stop producing incorrect pack,
> but there is no way other than sending a garbage string (as an
> error message) to cause pack-object to notice there is garbage
> coming in.

Ok, this patch should fix this.  Instead of "echo abort" another more
meaningful message could be used.  I checked the other scripts which
redirect stderr and they should be fine.  In the tests I found
t/t1200-tutorial.sh:38 
t/t1300-repo-config.sh:276
t/t5500-fetch-pack.sh:66
redirecting stderr, which probably fail which GIT_TRACE.

git-cvsserver.perl:2138 has 2>&1 but uses an regexp to match for the
content, so the trace message should not make a problem.  Can someone
using git-cvsserver confirm this?  Perhaps this should be changed to
2>/dev/null anyway.

---

 git-repack.sh |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/git-repack.sh b/git-repack.sh
index 640ad8d..4dd7dad 100755
--- a/git-repack.sh
+++ b/git-repack.sh
@@ -43,7 +43,7 @@ case ",$all_into_one," in
 	;;
 esac
 pack_objects="$pack_objects $local $quiet $no_reuse_delta$extra"
-name=$(git-rev-list --objects --all $rev_list 2>&1 |
+name=$( (git-rev-list --objects --all $rev_list || echo abort) |
 	git-pack-objects --non-empty $pack_objects .tmp-pack) ||
 	exit 1
 if [ -z "$name" ]; then
-- 
1.4.1.gf157-dirty

Re: [PATCH] fix git-repack for use with GIT_TRACE

[Junio C Hamano]

Matthias Lederhofer <matled@gmx.net> writes:

> Ok, this patch should fix this.  Instead of "echo abort" another more
> meaningful message could be used.

"git-rev-list died with exit code $?" perhaps.

> I checked the other scripts which
> redirect stderr and they should be fine.  In the tests I found
> t/t1200-tutorial.sh:38 
> t/t1300-repo-config.sh:276
> t/t5500-fetch-pack.sh:66
> redirecting stderr, which probably fail which GIT_TRACE.

Probably t/test-lib.sh should unset GIT_TRACE then like it does
for other environment variables.