cctbx wiki baptiste addition

Baptiste Carvello devel at baptiste-carvello.net
Tue Aug 7 16:46:33 UTC 2012


Le 07/08/2012 11:56, Picca Frédéric-Emmanuel a écrit :
> is it possible to have something like this in pseudo code
> 
> ok = True
> for test in tests:
>     ok &= test.run()
> 
> return ok

With the old test runner at least, it would be possible.

> another concern about this is that the log of the test should be sent to stdout to be present in the debian build log.
> if the log is only accessible from a file during the build, it is kind of useless...

This one should be OK. The log *is* sent to stdout. I assume upstream
use grep afterwards to check for failures, but the full log is there.

> ok and where are the data for this test ?
well, for that test, it is in subdirectories of
"cctbx_sources/fable/test", but I don't think there's a general rule.

> is it possible to create a cctbx-test-data package for this ?
I'd say we can include them in the cctbx-test package, unless you think
they are too big.

> licenses of the data ?
As "fable" is some sort of a compiler, they are more or less trivial
pieces of fortran code. No explicit license, but I guess the general
cctbx licence applies.

> wiki :)
indeed ;-)

> and if I correctly understand, thoses binaries are run from run_tests.py scripts.
> what is the commad line used to run all the tests ?

The binaries are just executed in a subprocess, some of them with arguments.

> We should imagine creating a cctbx-test package:
> - providing a /usr/bin/cctbx-test script with a very simple
>   man page explaining thaht cctbx-test execute all the cctbx tests.
> - put all the binary test program under
>   /usr/lib/cctbx/xxx this way no need to create a man page for all of them.

That's a great idea! That way we can run the same tests as upstream's
"run_tests.csh" script, with our installed files.

Cheers,
Baptiste



More information about the debian-science-maintainers mailing list