You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Daniel Shahaf <d....@daniel.shahaf.name> on 2010/07/14 13:55:21 UTC
splitted buildbot logs
Currently, the buildbot "testlog" outputs are quite large (12MB).
When a test fails, it could be more convenient for the developer inspecting the
buildbot log if the output was in smaller files (so we don't have to grep for
the /^FAIL/ inside a 12MB file *in the browser*).
Thus: http://pastebin.ca/1900435
What do you think about doing something like this in our buildbot instance?
Re: splitted buildbot logs
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Daniel Shahaf wrote on Thu, Jul 15, 2010 at 00:35:39 +0300:
> Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> > I think that ideally, one could click on a link, and see something like:
> >
> > [[[
> > Failed tests:
> > authz_test-1
> > basic_test-17
> > merge_test-582
> > ...
> >
> > Complete logs:
> > authz_tests
> > basic_tests
> > copy_tests
> > ...
> > ]]]
> >
>
> Does buildbot support a separate "failed tests" section?
>
Never mind; we could have a script copy the logs of the failed tests from the
big tests.log into a separate failed-tests.log. i.e., we'd have "stdio",
"testlog", and "testlog.fails-only" links.
(This approach is also better than the patch posted upthread.)
How to make this live? Is this something for each slave to configure for itself?
Daniel
Re: splitted buildbot logs
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> > Currently, the buildbot "testlog" outputs are quite large (12MB).
> >
> > When a test fails, it could be more convenient for the developer inspecting the
> > buildbot log if the output was in smaller files (so we don't have to grep for
> > the /^FAIL/ inside a 12MB file *in the browser*).
> >
> > Thus: http://pastebin.ca/1900435
> >
> > What do you think about doing something like this in our buildbot instance?
>
> Would this need to be deployed to each buildslave, or would it live on
> the server?
>
I don't know enough about the buildbot config to tell.
> I think that ideally, one could click on a link, and see something like:
>
> [[[
> Failed tests:
> authz_test-1
> basic_test-17
> merge_test-582
> ...
>
> Complete logs:
> authz_tests
> basic_tests
> copy_tests
> ...
> ]]]
>
Does buildbot support a separate "failed tests" section?
> Each of the lines (aside from the section headings) would be
> clickable, linking to respective test log output. We could even
> colorize the output in some way.
>
> Does buildbot support anything other than plain text for the log file link?
>
> -Hyrum
Re: splitted buildbot logs
Posted by Mark Phippard <ma...@gmail.com>.
On Wed, Jul 14, 2010 at 10:28 AM, Hyrum K. Wright
<hy...@mail.utexas.edu> wrote:
> On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
>> Currently, the buildbot "testlog" outputs are quite large (12MB).
>>
>> When a test fails, it could be more convenient for the developer inspecting the
>> buildbot log if the output was in smaller files (so we don't have to grep for
>> the /^FAIL/ inside a 12MB file *in the browser*).
>>
>> Thus: http://pastebin.ca/1900435
>>
>> What do you think about doing something like this in our buildbot instance?
>
> Would this need to be deployed to each buildslave, or would it live on
> the server?
>
> I think that ideally, one could click on a link, and see something like:
>
> [[[
> Failed tests:
> authz_test-1
> basic_test-17
> merge_test-582
> ...
>
> Complete logs:
> authz_tests
> basic_tests
> copy_tests
> ...
> ]]]
>
> Each of the lines (aside from the section headings) would be
> clickable, linking to respective test log output. We could even
> colorize the output in some way.
>
> Does buildbot support anything other than plain text for the log file link?
We had this for the Hudson builds, as well as timing trends.
--
Thanks
Mark Phippard
http://markphip.blogspot.com/
Re: splitted buildbot logs
Posted by "Hyrum K. Wright" <hy...@mail.utexas.edu>.
On Thu, Jul 15, 2010 at 10:29 AM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> Bert Huijben wrote on Thu, Jul 15, 2010 at 00:31:42 +0200:
>> (Should we add the XFail output to the failure log?)
>>
>
> Currently it does just FAIL and XPASS, rationale being that XFail is not an
> exceptional condition (we have some tests XFail for years).
+1
Re: splitted buildbot logs
Posted by 'Daniel Shahaf' <d....@daniel.shahaf.name>.
Bert Huijben wrote on Thu, Jul 15, 2010 at 00:31:42 +0200:
>
>
> > -----Original Message-----
> > From: Daniel Shahaf [mailto:d.s@daniel.shahaf.name]
> > Sent: woensdag 14 juli 2010 23:58
> > To: dev@subversion.apache.org
> > Cc: Daniel Shahaf
> > Subject: Re: splitted buildbot logs
> >
> > Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> > > On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> > > > Currently, the buildbot "testlog" outputs are quite large (12MB).
> > > >
> > > > When a test fails, it could be more convenient for the developer
> > > > inspecting the buildbot log if the output was in smaller files (so we
> > > > don't have to grep for the /^FAIL/ inside a 12MB file *in the
> > > > browser*).
> > > >
> > > > Thus: http://pastebin.ca/1900435
> > > >
> > > > What do you think about doing something like this in our buildbot
> > > > instance?
>
> Shouldn't this be handled by our python test runner instead?
>
> If that would just create an extra logfile next to the normal logfile with
> only the failing test as output, we would only have to reconfigure the
> master config to upload two instead of one file. Additional benefits are
> that it would help in local testing and that it would also work on
> non-posix platforms.
>
You persuaded me :) r964349.
Now just need to make buildbot upload the fails.log file too.
> (Should we add the XFail output to the failure log?)
>
Currently it does just FAIL and XPASS, rationale being that XFail is not an
exceptional condition (we have some tests XFail for years).
> Bert
>
> > >
> > > Would this need to be deployed to each buildslave, or would it live on
> > > the server?
> > >
> >
> > The former, per gmcdonald:
> >
> > 21:50 < gmcdonald> ok, so have the script run on the slave (becuase thats
> > where the logs are) then create an extra step on
> > the master to upload the results
> > 21:52 < gmcdonald> (the subversion.conf can add steps to run the script on
> > the slave (same as the cleanup step) then upload
> > the results to master)
> > 21:54 < gmcdonald> see also:
> > http://buildbot.net/buildbot/docs/0.8.1/full.html#BuildStep-LogFiles
> > 21:54 < gmcdonald> Using LogFiles in custom BuildSteps
> > 21:54 < gmcdonald> but whatever, you choose
>
Re: splitted buildbot logs
Posted by 'Daniel Shahaf' <d....@daniel.shahaf.name>.
Bert Huijben wrote on Thu, Jul 15, 2010 at 00:31:42 +0200:
>
>
> > -----Original Message-----
> > From: Daniel Shahaf [mailto:d.s@daniel.shahaf.name]
> > Sent: woensdag 14 juli 2010 23:58
> > To: dev@subversion.apache.org
> > Cc: Daniel Shahaf
> > Subject: Re: splitted buildbot logs
> >
> > Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> > > On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> > > > Currently, the buildbot "testlog" outputs are quite large (12MB).
> > > >
> > > > When a test fails, it could be more convenient for the developer
> > > > inspecting the buildbot log if the output was in smaller files (so we
> > > > don't have to grep for the /^FAIL/ inside a 12MB file *in the
> > > > browser*).
> > > >
> > > > Thus: http://pastebin.ca/1900435
> > > >
> > > > What do you think about doing something like this in our buildbot
> > > > instance?
>
> Shouldn't this be handled by our python test runner instead?
>
> If that would just create an extra logfile next to the normal logfile with
> only the failing test as output, we would only have to reconfigure the
> master config to upload two instead of one file. Additional benefits are
> that it would help in local testing and that it would also work on
> non-posix platforms.
>
You persuaded me :) r964349.
Now just need to make buildbot upload the fails.log file too.
> (Should we add the XFail output to the failure log?)
>
Currently it does just FAIL and XPASS, rationale being that XFail is not an
exceptional condition (we have some tests XFail for years).
> Bert
>
> > >
> > > Would this need to be deployed to each buildslave, or would it live on
> > > the server?
> > >
> >
> > The former, per gmcdonald:
> >
> > 21:50 < gmcdonald> ok, so have the script run on the slave (becuase thats
> > where the logs are) then create an extra step on
> > the master to upload the results
> > 21:52 < gmcdonald> (the subversion.conf can add steps to run the script on
> > the slave (same as the cleanup step) then upload
> > the results to master)
> > 21:54 < gmcdonald> see also:
> > http://buildbot.net/buildbot/docs/0.8.1/full.html#BuildStep-LogFiles
> > 21:54 < gmcdonald> Using LogFiles in custom BuildSteps
> > 21:54 < gmcdonald> but whatever, you choose
>
RE: splitted buildbot logs
Posted by Bert Huijben <be...@vmoo.com>.
> -----Original Message-----
> From: Daniel Shahaf [mailto:d.s@daniel.shahaf.name]
> Sent: woensdag 14 juli 2010 23:58
> To: dev@subversion.apache.org
> Cc: Daniel Shahaf
> Subject: Re: splitted buildbot logs
>
> Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> > On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name>
> wrote:
> > > Currently, the buildbot "testlog" outputs are quite large (12MB).
> > >
> > > When a test fails, it could be more convenient for the developer
> inspecting the
> > > buildbot log if the output was in smaller files (so we don't have to
grep for
> > > the /^FAIL/ inside a 12MB file *in the browser*).
> > >
> > > Thus: http://pastebin.ca/1900435
> > >
> > > What do you think about doing something like this in our buildbot
> instance?
Shouldn't this be handled by our python test runner instead?
If that would just create an extra logfile next to the normal logfile with
only the failing test as output, we would only have to reconfigure the
master config to upload two instead of one file. Additional benefits are
that it would help in local testing and that it would also work on
non-posix platforms.
(Should we add the XFail output to the failure log?)
Bert
> >
> > Would this need to be deployed to each buildslave, or would it live on
> > the server?
> >
>
> The former, per gmcdonald:
>
> 21:50 < gmcdonald> ok, so have the script run on the slave (becuase thats
> where the logs are) then create an extra step on
> the master to upload the results
> 21:52 < gmcdonald> (the subversion.conf can add steps to run the script on
> the slave (same as the cleanup step) then upload
> the results to master)
> 21:54 < gmcdonald> see also:
> http://buildbot.net/buildbot/docs/0.8.1/full.html#BuildStep-LogFiles
> 21:54 < gmcdonald> Using LogFiles in custom BuildSteps
> 21:54 < gmcdonald> but whatever, you choose
Re: splitted buildbot logs
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Hyrum K. Wright wrote on Wed, Jul 14, 2010 at 15:28:57 +0100:
> On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> > Currently, the buildbot "testlog" outputs are quite large (12MB).
> >
> > When a test fails, it could be more convenient for the developer inspecting the
> > buildbot log if the output was in smaller files (so we don't have to grep for
> > the /^FAIL/ inside a 12MB file *in the browser*).
> >
> > Thus: http://pastebin.ca/1900435
> >
> > What do you think about doing something like this in our buildbot instance?
>
> Would this need to be deployed to each buildslave, or would it live on
> the server?
>
The former, per gmcdonald:
21:50 < gmcdonald> ok, so have the script run on the slave (becuase thats where the logs are) then create an extra step on
the master to upload the results
21:52 < gmcdonald> (the subversion.conf can add steps to run the script on the slave (same as the cleanup step) then upload
the results to master)
21:54 < gmcdonald> see also: http://buildbot.net/buildbot/docs/0.8.1/full.html#BuildStep-LogFiles
21:54 < gmcdonald> Using LogFiles in custom BuildSteps
21:54 < gmcdonald> but whatever, you choose
Re: splitted buildbot logs
Posted by "Hyrum K. Wright" <hy...@mail.utexas.edu>.
On Wed, Jul 14, 2010 at 2:55 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> Currently, the buildbot "testlog" outputs are quite large (12MB).
>
> When a test fails, it could be more convenient for the developer inspecting the
> buildbot log if the output was in smaller files (so we don't have to grep for
> the /^FAIL/ inside a 12MB file *in the browser*).
>
> Thus: http://pastebin.ca/1900435
>
> What do you think about doing something like this in our buildbot instance?
Would this need to be deployed to each buildslave, or would it live on
the server?
I think that ideally, one could click on a link, and see something like:
[[[
Failed tests:
authz_test-1
basic_test-17
merge_test-582
...
Complete logs:
authz_tests
basic_tests
copy_tests
...
]]]
Each of the lines (aside from the section headings) would be
clickable, linking to respective test log output. We could even
colorize the output in some way.
Does buildbot support anything other than plain text for the log file link?
-Hyrum