Re: [jamboost] Patch for test-system to support testing in existing directories

From: Niklaus Giger (NiklausGiger_at_[hidden])
Date: 2005-03-15 13:54:53

Am Dienstag, 15. März 2005 17.28 schrieb Vladimir Prus:
> Hi Niklaus,
>
> > Here is my patch. On Cygwin there were about four tests that failed
> > before and after applying this patch.
> >
> > On my Powerbook (GNU/Linux with Debian testing) most Unittests fail
> > before and after after complaining that they could not find the gcc
> > toolset. I thought that getting a fresh copy from CVS, building bjam in
> > jam_src, then going to the v2/test-directory should result in a working
> > environment for tests. Or did I miss something. python test_all.py gcc
> > didn't help neither. Both gcc and g++ are installed and have version
> > 3.3.5.
>
> Did you add "using gcc ; " in user-config.jam, or in test-config.jam?
Oops. I had a user-config.jam in build/v2/test instead of build/v2.
> > Would it be helpful to setup a small daily cron job which fetches the
> > latest CVS, rebuilds bjam, runs the test and E-mails me the output, so
> > that I could forward it after the interesting aspects after a visual
> > inspection to this list? Is there somewhere such a script? (I cannot
> > guarantee that I would take a look at the output every day.)
>
> Victor is doing that to V2/msvc and that's rather helpful. What systems can
> you run tests at?
Cygwin (at work) and Debian GNU/Linux on PowerPC (at home).
> > Could please somebody revise this patch and commit if it is okay? Feel
> > free to suggest more improvements.
>
> Did you look at the patch? ;-) It touches every single line, so it's very
> hard to tell what you did modify. I guess you (or your editor), modified
> line ending for the file. But anyway, since I did not modified the files
> recently, it applies OK. I plan to apply the patch tomorrow.
Sorry for your inconvenience. I looked at it and corrected test_system.html, 
but forgot to convert also BoostBuild.py.

I just found out that on my PowerBook I had neither a TMP environment variable 
and that the output of `pwd` did not corresponde to the python os.getcwd, 
when I had symbolic link from /usr/src/boost -> /mnt/<somepath>. So my 
testcase failed miserably under these cicumstances.

Therefore I attach a better version of the patch and the test case here.

Thanks again for your help.

-- 
Niklaus Giger
Wieshoschet 6
CH-8753 Mollis
 --Boundary-00=_99yNCKivp9B5DMR Content-Type: application/x-gzip;
name="abs_dir.diff.gz"
Content-Transfer-Encoding: base64
Content-Disposition: attachment;
filename="abs_dir.diff.gz"
[Attachment content not displayed.] --Boundary-00=_99yNCKivp9B5DMR Content-Type: application/x-gzip;
name="abs_workdir.py.gz"
Content-Transfer-Encoding: base64
Content-Disposition: attachment;
filename="abs_workdir.py.gz"
[Attachment content not displayed.] --Boundary-00=_99yNCKivp9B5DMR--