log file should include branch information

[Barry Warsaw]

The log files written by pqm are less than helpful, especially if the branch does not complete successfully. The log file should include either the branch name or merge command right at the top, so grepping through a directory of log files can find the one you want. Currently, you have to guess, or look for a unique file from your branch, and this is not a good workaround.

[Robert Collins (lifeless)]

On Wed, 2008-06-11 at 15:05 +0000, Barry Warsaw wrote:
> Public bug reported:
>
> The log files written by pqm are less than helpful, especially if the
> branch does not complete successfully. The log file should include
> either the branch name or merge command right at the top, so grepping
> through a directory of log files can find the one you want. Currently,
> you have to guess, or look for a unique file from your branch, and this
> is not a good workaround.

From a recent merge (slightly edited to remove internal details):
====
Executing star-merge bzr+ssh://HOST/PATH at Wed Jun 11 21:07:53 2008
merge successful
Executing pre-commit hook /home/pqm/bin/kill-chroot-processes-hardy.sh
&& /home/pqm/bin/dchroot-run make TARGET VERBOSITY=-vvv at Wed Jun 11
21:08:15 2008
...
Test output here
pre-commit hook succeeded at Wed Jun 11 22:25:32 2008

star-merge succeeded at Wed Jun 11 22:25:33 2008
====

Clearly the log files have this.

Can you please expand on what you are referring to.

--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[Robert Collins (lifeless)]

On Wed, 2008-06-11 at 22:24 +0000, Robert Collins wrote:

> Clearly the log files have this.
>
> Can you please expand on what you are referring to.

I talked to Barry, and he pointed me at logs without this...

Something seems to be trashing older log files. I wonder if there is
some buggy script 'around' pqm that is mangling them?

-Rob
--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[Tom Haddon (mthaddon)]

No new log files have been written to the log directory for approx 13 hours. I'm not sure if this is related in any way to the log directory having over 16,000 files in it, or whether it's a PQM issue...

[Tom Haddon (mthaddon)]

As far as I'm aware, PQM has never included the branch information for when a merge fails. I've noticed this for quite a long time now, and had assumed that was just the way PQM worked.

[Barry Warsaw (barry)]

I think we can archive some of those old log files. It's painful to grep through so many anyway.

[Robert Collins (lifeless)]

On Thu, 2008-06-12 at 20:43 +0000, Tom Haddon wrote:
> As far as I'm aware, PQM has never included the branch information for
> when a merge fails. I've noticed this for quite a long time now, and had
> assumed that was just the way PQM worked.

I have checked the code and as long as it parses the request (that is,
it has a valid target, and what looks like it might be a valid source)
it will output the target branch before it actually attempts the merge.

-Rob
--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[Tom Haddon (mthaddon)]

I've archived the logs, but per my earlier message, PQM is still failing to write any log files to the logs directory. There's nothing in the pqm.log file that would indicate why it's failing to do this.

[Robert Collins (lifeless)]

Please confirm its still not logging; and assuming its not run a strace on the merge of a branch to find out what its doing, or put a pdb statement in where the output we want is and figure it out that way.

e.g. - disable pqm
wait for a merge
run the cronscript by hand with strace. You don't need follow children so it should be a relatively small strace. Or do the pdb thing (and still run by hand).

[Robert Collins (lifeless)]

ping?