Hi Juan,
On 6/6/10 7:35 PM, Juan Lang wrote:
Any comments and suggestions are welcomed.
One suggestion is that bugs get logged about test failures on Wine wherever possible. Dan has done that for the chromium test suite, and that's led to several bugs getting converted to Wine regression tests, and even a few fixes ;-)
I agree, that's something I'd like to see as well. There are, in fact, two ways interesting me for running these tests:
- Tests of MSVC builds
That would be similar Chrome tests. It would be compiled the same way (under Wine) as official Firefox, so we'd find Wine bugs this way. The work on it was done in parallel to Chrome work, but it was never completed. AFAIK the main stopper is currently cmd implementation, so the SoC project should help. Dan will know more details about it.
- Get tests of our mingw build
There is a config in Wine Gecko Git tree that enables building them. They compile fine, but there is a problem with running them. I've played a bit with it in the past. To run them, a combination of python and makefiles is used. The compilation is done in Linux environment, without Wine involved, so we have to teach Mozilla build environment to run tests under Wine. By adding a simple wrapper I was able to run some subset of tests and they found a few bugs. This way we test both problems with mingw compilation and Wine and they have to be separated. From my experience there were more build bugs than Wine's. They are, however, interesting for us as well (for me even more, as we can fix Wine anytime and Wine Gecko releases are rare). The problem with running more tests is that we really need python scripts to be ran under Windows python and that's more tricky. I've never found the time to investigate it further.
Cheers, Jacek