emerge --jobs>1, distcc, and different PORTAGE_TMPDIR ...

[eccerr0r]

I was wondering about this anomaly which has been vexing me, though it still seems to be "working" somewhat...

I'm currently building chromium and qtwebengine at the same time due to emerge --jobs=2 . I have PORTAGE_TMPDIR set to different directories with different environments.

However both emerges have separate DISTCC_DIRs because DISTCC_DIR is based off of PORTAGE_TMPDIR ? So basically, the two builds don't know about the jobs that the other build is doing as they have separate scoreboards.

I'm just trying to investigate why some of my distcc machines are actually idle while both merges are happening. My guess is that as usual, the load average of the job submitter is constantly getting exceeded but this doesn't appear to be the case.

Anyway I was trying to see what I can do with distcc and my new to me 10 core/20 thread/64GiB distcc helper that can definitely handle 20 huge c++ jobs at the same time.

--

hmm qtwebengine just finished and chromium is still building...

2024-07-08T04:28:27 >>> dev-qt/qtwebengine-5.15.14_p20240510: 2:38:46
2024-07-08T03:22:54 >>> www-client/chromium-126.0.6478.126: 3:48:41... (686 of 712) ETA: 2:18:51 (average run)

So far the additional cores does not look like it's helping much as both the distcc and distccd box have idle cores despite build.log seemingly having jobs ready to submit...

--
Okay this is probably the smoking gun, I see that chromium is failing a lot of distcc jobs:

distcc[84379] (dcc_build_somewhere) Warning: failed to distribute, running locally instead

maybe a helper somewhere that's killing the whole system...?

--

Chromium is now the last job of 712 that needs to complete and only now I'm seeing full CPU utilization...

--

and it finished.

2024-07-08T03:22:54 >>> www-client/chromium-126.0.6478.126: 7:18:50
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[Zucca]

I wonder if this is related...
_________________
..: Zucca :..

[eccerr0r]

Yeah, the job failures are probably related to that bug. Seems the llvm/clang versions are really not compatible with each other and mixing doesn't work. So to update machines with old llvm/clang I need to have the "fresher" machines retain the old llvm/clang to help out the machines that are updated later...

Unfortunately some machines disk space is at a premium...
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?