Consider using rustc-1.42.0

[ken@…]

If we can use rustc-1.42.0 we should be able to build it against system llvm-10.0, saving time and space. As always with rustc, firefox-ESR, thunderbird and seamonkey may require changes.

[ken@…]

For seamonkey, the link in #13271 to patches for 1.40 or 1.41 was [ ​​https://bugzilla.mozilla.org/show_bug.cgi?id=1617782 ]

[Pierre Labastie]

Built 1.42.0 successfully with system llvm-10.0.0. I logged measurements, but inadvertently scratched the report... Tests give:

• passed 18074
• failed 4

Will now try to build cbindgen and librsvg, then firefox and thunderbird. For seamonkey, I lack motivation: It has been broken during three months or so in the last release cycle, and nobody noticed...

[Pierre Labastie]

Ah, mozjs 68 needs rustc too, AFAIU. Will try this one too.

[Pierre Labastie]

Replying to pierre.labastie:

Ah, mozjs 68 needs rustc too, AFAIU. Will try this one too.

Eventually, it does not need rustc, thanks to xry111...

[Pierre Labastie]

librsvg and cbindgen build Ok. No change in test results.

[ken@…]

Thanks for looking at this.

I have not yet built it, will do that and take measurements (from 1.40 and 1.41 the tests results added up if you redefined what "run" means (I think it was something like run=passed+skipped), then look at seamonkey.

The big change in rust is usually size and time, but sometimes builds of the packages using it are faster with a new version.

[Pierre Labastie]

Firefox built Ok, too. A youtube video just playing now, no problem so far.

[Pierre Labastie]

Thunderbird built OK too. And connecting/retrieving mail seems Ok.

[ken@…]

Replying to pierre.labastie:

Built 1.42.0 successfully with system llvm-10.0.0. I logged measurements, but inadvertently scratched the report... Tests give:

• passed 18074
• failed 4

I guess you have gdb installed ?

For me, the total for 'running .* tests' is 17958, of which 17817 passed, 5 failed 141 ignored.

My failures are:

    [ui] ui/issues/issue-37131.rs
    [ui] ui/issues/issue-49851/compiler-builtins-error.rs
    [ui] ui/issues/issue-50993.rs
    [ui] ui/issues/issue-69225-SCEVAddExpr-wrap-flag.rs

    [run-make] run-make-fulldeps/sysroot-crates-are-unstable

That issue 69225 is the test for the memory unsafety problem fixed in 1.41.1 and for which I've been using the shipped llvm with 1.41.1.

double-plus-ungood. I checked that my llvm-10.0.0 in /opt was at the front of my path (script reported clang version 10.0.0).

[ken@…]

Also, something in my awk processing is not right, the overall total of tests run matches passed+ignored, but excludes failed. However when I look at the log for run-make-fulldeps that suite shows it is running 205 tests, then at the end that suite reports

test result: FAILED. 189 passed; 1 failed; 15 ignored; 0 measured; 0 filtered out

That does add up to 205, so the details need to be reviewed in detail at some point to find out where the discrepancy is coming from. Not urgent, unlike the 69225 failure.

[ken@…]

Replying to ken@…:

That issue 69225 is the test for the memory unsafety problem fixed in 1.41.1 and for which I've been using the shipped llvm with 1.41.1.

double-plus-ungood. I checked that my llvm-10.0.0 in /opt was at the front of my path (script reported clang version 10.0.0).

​https://github.com/rust-lang/rust/issues/69225

[Pierre Labastie]

I have:

    [ui] ui/issues/issue-37131.rs
    [ui] ui/issues/issue-49851/compiler-builtins-error.rs
    [ui] ui/issues/issue-50993.rs
[run-make] run-make-fulldeps/sysroot-crates-are-unstable

So no prob with 69225. I have gdb on this machine. Sorry, 18074 is the total number of tests run. The total number of passed is 17894 The total number of ignored is 181 With the 4 failed, it makes 5 tests difference...

[Pierre Labastie]

Ah, I see:

pierre [ ~/all-logs/upg-rust-1.42.0/work ]$ grep '^test result:' test-rust | wc -l
101
pierre [ ~/all-logs/upg-rust-1.42.0/work ]$ grep 'running.*tests' test-rust | wc -l
97

Ah , when running 1 test, test is singular, so not found by the regexp above... Unfortunately:

pierre [ ~/all-logs/upg-rust-1.42.0/work ]$ grep 'running.*test' test-rust | wc -l
253

This one seems ok (can it be simplified?):

pierre [ ~/all-logs/upg-rust-1.42.0/work ]$ grep 'running.*tests\?$' test-rust | wc -l
101

Eventually:

pierre [ ~/all-logs/upg-rust-1.42.0/work ]$ grep 'running.*tests\?$' test-rust | awk '{ sum += $2 } END { print sum }'
18079

[ken@…]

Thanks, I've now installed in /usr and the test for issue 69925 did not fail.

I'm now looking at my log because without gdb I got

17958 run
17816 passed
    4 failed
  143 ignored

which totals 17963, or 5 more.

As a reminder for myself, in 'view' I'm now using '/running\s\d\+\stests'to find the start of each batch, then adding them up. Hmm, maybe '/running\s\d\+\stest'.

[ken@…]

I'm coming to the opinion that the test results should be reworded to not count anything except the failures, to report expected failures, and to tell people that if they have other failures which mention an issue they should check it (69925 with a link would be useful as an example of an error thatshould be treated as faal, with explanation).

[Pierre Labastie]

Sounds good

[ken@…]

Thought I'd closed this last night from another machine, obviously I didn't 'submit'.

Done at r22919.

[Bruce Dubbs]

Milestone renamed

[Bruce Dubbs]

Milestone renamed