Firefox-62.0 slow to compile ?

[zerothruster]

I saw that the source for firefox-62.0 was out, so I downloaded. Previous version I've built was 62.0beta20. I use a patch for system harfbuzz and system graphite2, so the first thing I did was diff the sources to see if anything relevant to that or the other system versions had changed - no. Started the build, left it. From time to time checked, still building. Eventually used 'top', at that point 'ld' was running so it was only using 100% cpu and finished shortly after.

This is a Ryzen 1300X, with beta20 and -j4 the build and install had taken under 47 minutes. With 62.0 release, almost 131 minutes for ./mach build. Installed that, works fine, but thought I'd done something wrong in the build, or maybe --verbose was slowing it. Fresh source, dropped caches, retried without the patch : still building, not using more than 100% CPU (i.e. only one core) in either rustc or cc1plus, MOZ_MAKE_FLAGS=-j4 definitely exported.

LFS/BLFS-8.3, gcc-8.2.0, llvm/clang 6.0.1, rustc-1.25.0. Kernel, if it matters, is 4.18.3 (and all are unchanged since I built beta20).

[princemoga]

Firefox team maybe help you..

[zerothruster]

Maybe, but like most bugs it will probably just sit around unless it happens to a lot of people. It did get moved from Untriaged and Firefox to General and Firefox Build System.

I suspect I might have seen similar odd behaviour in the past, but I might have been thinking of when I was compiling rust. Both use python for the build system, both seem to often be erratic. My experience when building both of them in scripts is that they can fail because of errors, but somehow return a status of success (i.e. true) to a script which invoked them.

[Bates357]

Surely, Its very obvious that, the new version of firefox is very slow to compile it with other softwares. This is probably happens because when you start downloading a new version there may be a bug that creates certain error to file and thus it needs more and more time to get compiled. So, Just service your device and attach it with good antivirus protection.

[duby229]

Surely, Its very obvious that, the new version of firefox is very slow to compile it with other softwares. This is probably happens because when you start downloading a new version there may be a bug that creates certain error to file and thus it needs more and more time to get compiled. So, Just service your device and attach it with good antivirus protection.

What? Say that again. I don't understand what you mean.

[Bates357]

Surely, Its very obvious that, the new version of firefox is very slow to compile it with other mobdro softwares. This is probably happens because when you start downloading a new version there may be a bug that creates certain error to file and thus it needs more and more time to get compiled. So, Just service your device and attach it with good antivirus protection.

I mean, I also feel that Firefox-62.0 is slow to compile.

[duby229]

I mean, I also feel that Firefox-62.0 is slow to compile.

OK sure. But explain why you think it might be caused by corrupt download? I'm confused by that and why you think a antivirus could help?

[zerothruster]

Meanwhile, I've now seen similar issues with 62.0.2 and 63.0b8 - but only on the Ryzen 1300X. The build of 62.0.2 was mostly very slow (fell back to only 100% CPU), but eventually speeded up to using 400% near the end. With 63 I've been using rustc-1.29.0, first build (clang) was very slow but eventually failed (possibly something in my choice of system libs). Went back to gcc-8.2.0, it was back to 400% a lot of the time.

I still think the (python) build system is unpredictable, but at least with both rustc and firefox-63 compile failures seem to propagate out and stop my build, which is much better than how things were a few months ago.

[zerothruster]

Since people (or perhaps, just insane attempts at artificial intelligence) are still posting to this old thread, what I hope will be a final comment from me: For building from source, rustc has horrendous variability - even when rebuilding the same package (firefox or whatever) on the same machine and same system, so that all the cargo files have already been downloaded, the timings are all over the place. Sometimes, you get a faster build. Other times, you get a very slow build. It's just the nature of the beast.

All rust really aims to do is to provide safer builds, and for that aim, or hope, it's far too early to form a view on whether or not it meets it.

[nblinks]

I think the firefox support team are in the best position to solve this.