qtwebengine-6.7.2 chromium version compile issuesFIXED

[wrc1944]

qt/qtwebengine-6.7.2:6/6.7.2 always fails about 2/3rds of the way through, even after seriously enlarging my gentoo and swap partitions with gparted,as the /var/tmp/dev-qt/ build environment file suggested would likely cause compile failure.' and it did, but since I have 6.7.1 version installed this must be a new issue.



I suspect the newest google-chrome-126.0.6478.114 version is still the main problem,as, while building is throws countless deprecations like this:

[eccerr0r]

The excerpt does not contain the information needed to debug. The deprecations are warnings and should not affect build, else it would fail on the first one.

Please post the entire build.log as well as the output of "emerge --info qtwebengine" somewhere and link it to here (pastebin, dpaste, etc., etc.)

I'm not sure what -l0 does with ninja. This according to help would not be very useful. Use -l1 at minimum and nothing lower. In any case it's the -j option that has any affect on memory consumption, as far as I can tell, setting -l0 just slows things down without much upside.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[wrc1944]

eccerr0r,
Thanks for the reply!



Had a hard time with paste bin, so tried this.Only got first part of the build log. I'll try again if this isn't enough. Will post other stuff as I figure it out.

Please let me know if the little dpaste was of any use.


https://dpaste.com/CTKJBYD6Y


To clear the opts I used in make.conf, I only tried =J1, =J4, and =J8, and No others, and all three failed in the same manner. Not sure where the J10 came from.


Here's the other info you mentioned.

[Ionen]

I have an idea what might have happened, likely xnnpack bits failing due to the qt6-build.eclass needing to handle -mno-xop being passed

.. but I still need the full build.log to be able to confirm what happened, the error is not included in either of these pastes.

[wrc1944]

I searched the build file and close to the end found many xnnpack items before it failed. maybe this wikk help?



https://dpaste.com/ELAHSM7PY
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.39-r7, gcc-14
kernel-6.9.4 USE=experimental python3_12.3-final

[Ionen]

Thanks, that's what I was expecting (that's xop intrinsics). So I've done https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=ab3a604aa77a to handle this.

Please wait a bit (~15-30 mins usually) for changes to hit mirrors then emerge --sync, hopefully it'll be enough.

To confirm that the workaround is in effect, you should see something like:

[Ionen]

On another note, combining -march=native with -mno-* flags is kind of odd. native should auto-detect everything properly.

This is what caused your problems here, and ideally you should remove these. Not that I didn't need to work around this either way for normal use cases (aka without =native). If you only use "-march=native -O2 -pipe" then the workaround line won't appear and it should also build fine anyway.

Edit: By the way, -fomit-frame-pointer is default with -O2 on x86-64, so you don't need to pass it either.

[wrc1944]

lonen,

Thanks very much!
On the -march=native with -mno-* flags I think I had to have that cflags line as it was reportedly needed for the early amd Ryzen 7 cpu.

Might be wrong on that, it was many years ago.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.39-r7, gcc-14
kernel-6.9.4 USE=experimental python3_12.3-final

[wrc1944]

lonen,
Just to be sure, what needs to be changed in my CFLAGS line in make.conf?

-pipe -fomit-frame-pointer -mno-fma4 -mno-tbm -mno-xop -mno-lwp"
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.39-r7, gcc-14
kernel-6.9.4 USE=experimental python3_12.3-final

[Ionen]

Just the default of COMMON_FLAGS="-march=native -O2 -pipe" should be fine, not super familiar with early ryzens but I don't imagine there being any problems given gcc will auto-detect, unless some usable instructions are straight up buggy somehow.

[wrc1944]

Thanks much, lonen, worked out fine with opts =J2, even if the last half of the compile was very,very slow, and paused many times for10-20 minutes leaving me wondering if thjis is yet another lock-up and failure.


Figure this must be lots of swapping going on, and I need 32Gb of fast RAM instead of my current 16Gb.

However, before qtwebengine-6.7.2 I had compiled previous versions right along in under 3 hours with opts set =J8..
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.39-r7, gcc-14
kernel-6.9.4 USE=experimental python3_12.3-final

[Ionen]

16GB is plenty to use -j6 w/ qtwebengine (could do a bit more but to be safe) as long as you don't waste ram on something else, like PORTAGE_TMPDIR being on tmpfs.