Re: [PATCH] rebase -i: fix misleading error message after 'exec no-such' instruction
Johannes Sixt writes: > From: Johannes Sixt > > When the todo sheet of interactive rebase instructs to run a non-existing > command, the operation stops with the following error: > > Execution failed: no-such > You can fix the problem, and then run > > git rebase --continue > > fatal: 'rebase' appears to be a git command, but we were not > able to execute it. Maybe git-rebase is broken? > > The reason is that the shell that attempted to run the command exits with > code 127. rebase--interactive just forwards this code to the caller (the > git wrapper). But our smart run-command infrastructure detects this > special exit code and turns it into ENOENT, which in turn is interpreted > by the git wrapper as if the external command that it just executed did > not exist. That's a funny one ;-) Will queue. Thanks. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] rebase -i: fix misleading error message after 'exec no-such' instruction
Johannes Sixt writes: > Am 9/18/2012 13:20, schrieb Matthieu Moy: >> Johannes Sixt writes: >> >>> From: Johannes Sixt >>> >>> When the todo sheet of interactive rebase instructs to run a non-existing >>> command, the operation stops with the following error: >>> >>> Execution failed: no-such >>> You can fix the problem, and then run >>> >>> git rebase --continue >>> >>> fatal: 'rebase' appears to be a git command, but we were not >>> able to execute it. Maybe git-rebase is broken? >> >> While you're there, maybe you want to turn the first line into >> >> Execution failed: no-such (command not found) > > No, I don't want to: Neither do we have errno here, nor can we be specific > enough because the whole shell script the user gave after 'exec' is > repeated here. > > I would rather remove the line so that it does not distract from the more > specific error message that the shell gave. On top of that, the command is > already dumped before it is executed; we don't need to repeat it. Not sure what you mean: remove the line in case $status = 127, or remove it anyway. If you mean only when $status = 127, then that is indeed a good idea, as the full error message looks like Executing: nosuchcommand zsh:1: command not found: nosuchcommand Execution failed: nosuchcommand You can fix the problem, and then run [...] So, yes, the shell's error message is enough and more precise than we could do in git rebase. If you mean remove it in any case, I disagree: repeating the command is usually not very useful, but may help when the command itself produced a lot of output (e.g. "exec make test" oftens fills-in your terminal's buffer). And the information that the command fails can be important if the command failed silently, e.g. with "exec false": Executing: false Execution failed: false You can fix the problem, and then run [...] -- Matthieu Moy http://www-verimag.imag.fr/~moy/ -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] rebase -i: fix misleading error message after 'exec no-such' instruction
Am 9/18/2012 13:20, schrieb Matthieu Moy: > Johannes Sixt writes: > >> From: Johannes Sixt >> >> When the todo sheet of interactive rebase instructs to run a non-existing >> command, the operation stops with the following error: >> >> Execution failed: no-such >> You can fix the problem, and then run >> >> git rebase --continue >> >> fatal: 'rebase' appears to be a git command, but we were not >> able to execute it. Maybe git-rebase is broken? > > While you're there, maybe you want to turn the first line into > > Execution failed: no-such (command not found) No, I don't want to: Neither do we have errno here, nor can we be specific enough because the whole shell script the user gave after 'exec' is repeated here. I would rather remove the line so that it does not distract from the more specific error message that the shell gave. On top of that, the command is already dumped before it is executed; we don't need to repeat it. -- Hannes -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] rebase -i: fix misleading error message after 'exec no-such' instruction
Johannes Sixt writes: > From: Johannes Sixt > > When the todo sheet of interactive rebase instructs to run a non-existing > command, the operation stops with the following error: > > Execution failed: no-such > You can fix the problem, and then run > > git rebase --continue > > fatal: 'rebase' appears to be a git command, but we were not > able to execute it. Maybe git-rebase is broken? While you're there, maybe you want to turn the first line into Execution failed: no-such (command not found) In any case, the patch sounds good, thanks. -- Matthieu Moy http://www-verimag.imag.fr/~moy/ -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH] rebase -i: fix misleading error message after 'exec no-such' instruction
From: Johannes Sixt When the todo sheet of interactive rebase instructs to run a non-existing command, the operation stops with the following error: Execution failed: no-such You can fix the problem, and then run git rebase --continue fatal: 'rebase' appears to be a git command, but we were not able to execute it. Maybe git-rebase is broken? The reason is that the shell that attempted to run the command exits with code 127. rebase--interactive just forwards this code to the caller (the git wrapper). But our smart run-command infrastructure detects this special exit code and turns it into ENOENT, which in turn is interpreted by the git wrapper as if the external command that it just executed did not exist. This is finally translated to the misleading last two lines in error message cited above. Fix it by translating the error code before it is forwarded. Signed-off-by: Johannes Sixt --- An alternative fix would be to just ignore the shell's status code. I decided that it is worth keeping it: better safe than sorry. BTW, the problem existed since day 1 of the 'exec' feature. git-rebase--interactive.sh| 4 t/t3404-rebase-interactive.sh | 11 +++ 2 files changed, 15 insertions(+) diff --git a/git-rebase--interactive.sh b/git-rebase--interactive.sh index a09e842..56707d7 100644 --- a/git-rebase--interactive.sh +++ b/git-rebase--interactive.sh @@ -544,6 +544,10 @@ do_next () { warn warn " git rebase --continue" warn + if test $status -eq 127 # command not found + then + status=1 + fi exit "$status" elif test "$dirty" = t then diff --git a/t/t3404-rebase-interactive.sh b/t/t3404-rebase-interactive.sh index 7304b66..7a71760 100755 --- a/t/t3404-rebase-interactive.sh +++ b/t/t3404-rebase-interactive.sh @@ -118,6 +118,17 @@ test_expect_success 'rebase -i with the exec command checks tree cleanness' ' git rebase --continue ' +test_expect_success 'rebase -i with exec of inexistent command' ' + git checkout master && + test_when_finished "git rebase --abort" && + ( + FAKE_LINES="exec_this-command-does-not-exist 1" && + export FAKE_LINES && + test_must_fail git rebase -i HEAD^ >actual 2>&1 + ) && + ! grep "Maybe git-rebase is broken" actual +' + test_expect_success 'no changes are a nop' ' git checkout branch2 && git rebase -i F && -- 1.7.12.1720.gb55457a.dirty -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html