Re: [scons-devel] First split up of SConf.py
Christoph Wiedemann <[email protected]>
| Newsgroups | gmane.comp.tools.aap.devel |
|---|---|
| Message-ID | <[email protected]> |
Hi Bram, > At the end you will find a diff for splitting up the SConf.py file into > a part that is SConf-specific and a generic part, called Conftest.py. > > The generic part is also used by the configure feature of Aap. > I am currently checking in the changes into CVS (see the download page: > http://www.a-a-p.org/download.html). Note that SourceForge has a backup > CVS server, it may not have the new version 1.012 for a while. Make > sure you use the real-time server for downloading. > > The goal was to put the parts of the configure checks that do not depend > on the build system in Conftest.py. To make this possible an object is > passed to the check functions that has to offer specific methods and > members. This is documented at the start of the Conftest.py file. > Together with the public functions of Conftest.py this forms the > interface to the generic configure checks. > > The functionality of SCons is unchanged. I have run all the tests I > could find. The "test/Configure.py" test had to be adjusted, because > the stdout of the tests changed slightly. > > For Aap the new ":conf" command has been implemented. Some of this code > is from Joerg Beyer. There is no documentation yet, sorry! Look in > DoConf.py if you want to see how it works. > > A few remarks: > > - Renamed "show_result" to "did_show_result" to make clear what the flag > actually stands for. > > - The tests can define HAVE_ABC variables and append them to a file > (normally called confdefs.h). This is required for many tests to > work, but since SCons didn't do this before, it is not done now > either. Aap does use this feature. > > - I added comments about why a test is done in the way it is done. This > functions as a knowledge base, so that people who change the tests in > the future understand the reasoning behind the test method. > > - Tests displayed "failed" when the test successfully detected that a > feature is not present. I find this confusing, thus changed it into > "yes". "failed" should be used when the test could not determine if > the feature is present or not. > > - The SConf.py function arguments are not consistent, I have added a few > comments to point this out. > This looks really well. I've just stepped quickly through the changes, and i like this patch. You've improved much of the test-docs, which is really useful. In fact, we should force developers to document the tests in this way, so everybody knows why something is made in a certain way. Thanks for all the work. Christoph ------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01