On Tue, May 26, 2015 at 2:14 PM, Dylan Baker <baker.dyla...@gmail.com> wrote: > On Tue, May 26, 2015 at 02:05:41PM -0400, Ilia Mirkin wrote: >> On Tue, May 26, 2015 at 2:06 PM, Dylan Baker <baker.dyla...@gmail.com> wrote: >> > On Tue, May 26, 2015 at 11:15:36AM +0100, Jose Fonseca wrote: >> >> argparse type is a nice type saver for simple data types, but it doesn't >> >> look a good fit for the input XML file: >> >> >> >> - Certain implementations of argparse (particularly python 2.7.3's) >> >> invoke the type constructor for the default argument even when an >> >> option is passed in the command line. Causing `No such file or >> >> directory: 'gl_API.xml'` when the current dir is not >> >> src/mapi/glapi/gen. >> >> >> > The fix is fine, I'm not complaining about the changes. >> > >> > However, >> > >> > I'm running 'scons' against python 2.7.3 and it's building fine with >> > your patch reverted. Is there something I'm missing about the way that >> > the scons build works? The documentation says that just 'scons' is all >> > that is necessary, is that not the case? >> >> Do an out-of-tree build where gl_API.xml isn't in the current directory. >> > > out-of-tree builds were not brought up in the discussion, I have been > working under the assumption of in-tree-builds. > > I'm also not sure at all how that's relavent, since the scons build > doesn't pass absolute paths, so falling back to the default or passing > '-f' should come up with the same result. > > Or did I completely misunderstand what you're saying?
Not *completely* :) So yes, you're right, out-of-tree builds should work. But note my analysis in the bug. Basically argparse will evaluate the default value *even if* the argument is overridden (for older versions of argparse). So it'll try to load 'gl_API.xml', which will either be there or not depending on the cwd of the command. When it's an in-tree build, the cwd is the gen dir which has the xml file and all's well. When it's an out-of-tree build, the cwd is in your side-build tree, and there's no gl_API.xml file there. At least that's my understanding of the situation. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
