dejagnu broken (at least according to gcc45)

[Ryan Schmidt]

On Apr 19, 2011, at 19:23, Jack Howarth wrote:
>   Actually it seems to have something to do with executing dejagnu in the ports build directory.
> In my account, I have a ~/ports/lang/gcc45 which creates a work directory within the local gcc45
> port. I don't understand the rules for creating the subdirectories in work during port file builds.
> I've found instances where ~/lang/gcc45/work will exist during a build but will appear to be empty
> whereas the complete path to the actual build will be shown by -d. I assume port is doing some 
> funky stuff with symlinks for the gcc45 build directory path and that might be confusing dejagnu.

It's nothing funky; $(port dir gcc45)/work is simply a symlink to the build directory. That's all there is to it. The symlink is only there for your convenience; MacPorts itself doesn't use it. If you don't want this symlink, you can tell MacPorts not to create it, by editing macports.conf; the setting is called place_worksymlink. It's described here:

http://guide.macports.org/#internals.configuration-files.macports-conf

Curiously, this option is not mentioned in the comments in macports.conf and I'm not sure what the reason is for this omission.


> I just do a manual build of gcc from the sources, dejagnu works fine. Is there a rule for where
> the real path to the build is expected to reside (with resorting to any symlinks in the path)?
> I mean like /sw/src/fink.build where the build directories reside in fink without any intervening
> symlinks in the path.

The build directory's actual location is printed by the command: port work gcc45