You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openoffice.apache.org by Gavin McDonald <gm...@apache.org> on 2015/05/08 10:03:26 UTC

[BUILDBOT] - Linux 32bit Nightly (openoffice-linux32-nightly)

Hi All,

So this build was ‘flapping’, i.e. passing/failing nearly every other build.

http://ci.apache.org/builders/openoffice-linux32-nightly?numbuilds=50 <http://ci.apache.org/builders/openoffice-linux32-nightly?numbuilds=50>

This was mainly due to the use of the method=‘copy’ way of updating the source tree.

The SVN step performs two things when in ‘copy’ mode, it does a checkout to a ‘source’
directory and then does a copy from there to the ‘build’ directory. The first has a default
timeout of 1200 whilst the latter has a (currently) unconfigurable timeout of 120.
It is the latter 120 timeout during copy that makes the build fail, about 50% of the time it
hits the timeout limit before the copy is complete and so the build fails.

I have changed from using "mode=‘full', method=‘copy’” to using "mode=‘incremental’”
which basically means use ‘svn update’ directly on the build directory. I believe this will
stabilise this particular build. (I can’t remember why mode=copy was used originally.)

All the other builds also still use mode=copy but I haven’t spotted a failing pattern with
those so far because of it.

HTH

Gav…


Re: [BUILDBOT] - Linux 32bit Nightly (openoffice-linux32-nightly)

Posted by jan i <ja...@apache.org>.
On 8 May 2015 at 10:03, Gavin McDonald <gm...@apache.org> wrote:

>
> Hi All,
>
> So this build was ‘flapping’, i.e. passing/failing nearly every other
> build.
>
> http://ci.apache.org/builders/openoffice-linux32-nightly?numbuilds=50
>
> This was mainly due to the use of the method=‘copy’ way of updating the
> source tree.
>
> The SVN step performs two things when in ‘copy’ mode, it does a checkout
> to a ‘source’
> directory and then does a copy from there to the ‘build’ directory. The
> first has a default
> timeout of 1200 whilst the latter has a (currently) unconfigurable timeout
> of 120.
> It is the latter 120 timeout during copy that makes the build fail, about
> 50% of the time it
> hits the timeout limit before the copy is complete and so the build fails.
>
> I have changed from using "mode=‘full', method=‘copy’” to using
> "mode=‘incremental’”
> which basically means use ‘svn update’ directly on the build directory. I
> believe this will
> stabilise this particular build. (I can’t remember why mode=copy was used
> originally.)
>
years ago, there was a problem with "svn up", that some files were not
correctly updated (something to do
with the filesystem, if I remember right).

In my opinion "svn up" should do just fine now (I use it on ubuntu and
windows without problems).

rgds
jan I.



> All the other builds also still use mode=copy but I haven’t spotted a
> failing pattern with
> those so far because of it.
>
> HTH
>
> Gav…
>
>