According to ABCD in IRC, I'm filing this.
Basically, kmail failed to compile giving:
make: *** No rule to make target `kmail/org.kde.kmail.kmail.xml', needed by `kontact/plugins/kmail/kmailinterface.cpp'. Stop.
When I add "-j1" to the MAKEOPTS in /etc/portage/env/kde-base/kmail, kmail completes successfully. Ideally, the dependencies would get fixed upstream, but, in the meantime, adding "-j1" to the make command in kmail may be sufficient.
There is a temporary fix (-j1) in the ebuild, and a bug has been filed upstream at <https://bugs.kde.org/show_bug.cgi?id=198807>.
I took out the -j1 option from my kmail-4.3.3 ebuild, and rebuilt it. It rebuilt fine. It looks like upstream did fix this bug between 4.2.95 and 4.3.3.
(I'd mark this bug fixed, but I think that we should now be able to remove the -j1 flag before closing it.)
I just checked, and it still can fail with current trunk (that doesn't mean it will always fail, or even will usually fail, but it *can* fail due to underspecified dependencies).
If this is only a compile problem, we can add an ewarn in the ebuild. But this can be done if this is only a compile problem, not if it also causes any runtime issues. Kmail takes a long time to compile now, so i guess it would be the best option atm, since upstream doesn't seem to care that much about this
As far as I can tell, it's a compile-time only issue. The main problem is that if make tries to build certain files in kontact/plugins/kmail/ before other files in kmail/, it fails (and there is no way that I am aware of to create the proper dependencies in cmake, unfortunately).
Seems to work fine with my -j9 on kde 4.4.2.
But then, this is a non-deterministic bug class... :)
(In reply to comment #6)
> Seems to work fine with my -j9 on kde 4.4.2.
> But then, this is a non-deterministic bug class... :)
Did you edit the ebuild to remove the "-j1" bit - otherwise, you *didn't* test with -j9. (I think this may have been fixed for 4.5, but I haven't actually checked yet)
(In reply to comment #7)
> Did you edit the ebuild to remove the "-j1" bit - otherwise, you *didn't* test
> with -j9. (I think this may have been fixed for 4.5, but I haven't actually
> checked yet)
In the beginning, no. Now I did it...
Worked fine for kmail-4.4.2 two times in a row with -j9
(and top really displayed the concurrent compilers... :)
I commented out the "-j1" lines in the ebuild for 4.4.4. When this moves to the tree, we'll get some more testing...
Nobody screaming so let's assume this is fixed.