dev-qt/qtwebengine-5.7.1-r2 failed to build anything.

Lebkoungcity

Greetings everyone,

I'm about to recompile my systems after switching to profile 17. Unfortunately qtwebengine failed after about one day and a half of compiling (btw: Is this long time a consequence of profile 17?). I still have the files in /var/tmp/portage and so I tried it once again with FEATURE="keepwork". But it still fails with

Hu · Moderator Joined: 06 Mar 2007 Posts: 21605

That message appears if ${D%/}${QT5_LIBDIR}/libQt5WebEngine.so fails [[ -f ]]. Based on your output, that name is a link, not a file. Even so, a link to a file should pass that test if the target exists and is a file. With the caveat that I do not maintain this package, this looks to me like a bug in the ebuild. This was added for dev-qt/qtwebengine-5.6.2 emerges despite unmet deps, but installs no files by kensington in June, so it seems strange that someone would just now hit this. Regardless, this looks to me to be a completely incorrect solution. If the dependencies are unmet, the compile ought to fail outright and the ebuild then die because the compile fails. Testing afterward for specific files is an ugly hack. Based on your experience, the hack is fragile, too.

[Edit to add a clarification that [[ -f ]] treats a link to a file as a file.]

kensington

Lebkoungcity

Chiitoo · Posted: Tue Dec 12, 2017 5:51 pm Post subject: ><)))°€

Something like this should do it:

Lebkoungcity · Posted: Tue Dec 12, 2017 6:36 pm Post subject:

OK, got it:

Hu · Moderator Joined: 06 Mar 2007 Posts: 21605

Lebkoungcity · Posted: Wed Dec 13, 2017 8:45 pm Post subject:

So, do I understand this correctly if I assume that this patch would let the building die early instead of letting it die in the end? That would have been nice compared to a long lasting compile process leading to a non-working result.

But is there a chance to get it successfully on my little machine?

edit: typo
_________________
"The most dangerous world view is the world view of the ones who haven't viewed the world."
Alexander von Humboldt (1769-1859)

Hu · Moderator Joined: 06 Mar 2007 Posts: 21605

If the patch, which I wrote with no prior experience with qmake and only a few minutes reading their documentation, works as intended, then yes, it would move real failures up to before the compile runs and would avoid false failures as you seem to experience. As I don't use the package, I did not test the patch for correctness in either the success or failure cases, but instead posted it straight away, per kensington's invitation to improve the logic. It's very possible that the patch will need refinement by someone more familiar with qmake.

It is probably possible to make this package work on your machine, but I don't know how to do so. You showed that the link exists, which is good. Does it point to a valid file? What is the output of ls -l /var/tmp/portage/dev-qt/qtwebengine-5.7.1-r2/image/usr/lib/?

Chiitoo · Posted: Thu Dec 14, 2017 9:19 am Post subject:

Indeed, things seem to have been built, but why the file is under '/var/tmp/portage/dev-qt/qtwebengine-5.7.1-r2/image/' instead of '.../usr/lib/' is a mystery that requires some more investigation.

I'm not sure there is much point in this, but if you like, you could try adding

kensington · Posted: Thu Dec 14, 2017 9:31 am Post subject:

Lebkoungcity · Posted: Thu Dec 14, 2017 6:01 pm Post subject:

Thanks a lot for your input!

Chiitoo · Posted: Thu Dec 14, 2017 6:26 pm Post subject:

Hm.

Does '/usr/include/qt5/Gentoo' really not exist (at least I think that's what it's looking for)?

To my understanding, at least 5 Qt packages install there, starting from 'qtcore'.

I believe it should contain something like this:

Lebkoungcity · Posted: Thu Dec 14, 2017 7:57 pm Post subject:

Chiitoo · Posted: Thu Dec 14, 2017 8:24 pm Post subject:

Thanks!

I'm not sure if the '/Gentoo/gentoo-qconfig.h' should be an absolute path that should in fact be '/usr/include/qt5/Gentoo/gentoo-qconfig.h'... and why it would not be that.

Hmmm...
_________________
Kindest of regardses.