Solved Firefox, Synth or not and LTO

[Geezer]

If you use poudriere, you'll always have a complete log

Synth has got logs too. Having rebuilt the same ports again without LTO, the logs have been overwritten.

Anyway, I do not believe that LTO can be a big benefit...

Yup.

 

[Styrsven]

FYI, thunderbird failing with LTO is already reported.
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=255311

 

[grahamperrin]

… Firefox with LTO. …

FreeBSD bug 257274 – www/firefox: Fails to link on i386 with LTO=ON on i386: ld: error: undefined hidden symbol: mozilla::ServoTraversalStatistics::sActive (firefox-90.0,2)

Retrospective: www/firefox: enable LTO by default on amd64 (2021-05-11) and other LTO-related commits.

 

[grahamperrin]

… to get a reliable reading seems hard. I have not seen any e-mail client benchmarks…

Benchmarks aside, we can probably look forward to Thunderbird becoming faster.

Thunderbird 91 Available Now | The Thunderbird Blog (2021-08-12)

(There's not yet a bug for version 91 or greater to be ported.)

 

[Geezer]

I have changed the subject of this thread ... yet again.

 

[Argentum]

FYI, thunderbird failing with LTO is already reported.
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=255311

And I can confirm this on an independent installation. The build ends with:

ld: error: undefined hidden symbol: JS::CleanupQueuedFinalizationRegistry(JSContext*, JS::Handle<JSObject*>)
>>> referenced by Unified_cpp_xpcom_base0.cpp
>>>               lto.tmp:(mozilla::CleanupFinalizationRegistriesRunnable::Run())

ld: error: too many errors emitted, stopping now (use -error-limit=0 to see all errors)
clang++: error: linker command failed with exit code 1 (use -v to see invocation)

Many undefined symbols before that...

 

[grahamperrin]

258536 – Mk/bsd.lto.mk: Add Link-Time Optimization (LTO) support

bugs.freebsd.org

 

[Oleg_NYC]

Speaking of LTO, I encountered something bizarre after enabling it in the configuration menu for the libreoffice port and keeping default values for this port's other configuration options. I compiled the port, but libreoffice crashed every time I tried to start it. Then I added WITH_DEBUG=yes to poudriere's make.conf, and, while keeping the LTO configuration option enabled, I re-compiled the port... And this time there were no crashes when I was running libreoffice. I have no idea why WITH_DEBUG=yes resulted in LTO-enabled libreoffice that didn't crash. By the way, WITH_DEBUG=yes still resulted in a stripped soffice.bin binary! This is all so bizarre and mysterious.

 

[astyle]

Speaking of LTO, I encountered something bizarre after enabling it in the configuration menu for the libreoffice port and keeping default values for this port's other configuration options. I compiled the port, but libreoffice crashed every time I tried to start it. Then I added WITH_DEBUG=yes to poudriere's make.conf, and, while keeping the LTO configuration option enabled, I re-compiled the port... And this time there were no crashes when I was running libreoffice. I have no idea why WITH_DEBUG=yes resulted in LTO-enabled libreoffice that didn't crash. By the way, WITH_DEBUG=yes still resulted in a stripped soffice.bin binary! This is all so bizarre and mysterious.

Yeah, that is weird - I always turn the debugging option off when I compile my ports, and I have yet to see that being a problem. Well, some ports still fail to compile for me, so that's an option to try next time...

 

[Alain De Vos]

If have seen no performance improvement at all using LTO.