1213165 - IonMonkey: MIPS32: Fix ion/bug901086.js failed

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[hev [:hev]]

[heiher@lemote jit-test]$ /tmp/js.32 -f /mnt/git/mips64/gecko-dev/js/src/jit-test/lib/prologue.js --js-cache /mnt/git/mips64/gecko-dev/js/src/jit-test/.js-cache -e "const platform='linux2'; c
onst libdir='/mnt/git/mips64/gecko-dev/js/src/jit-test/lib/'; const scriptdir='/mnt/git/mips64/gecko-dev/js/src/jit-test/tests/ion/'" -f /mnt/git/mips64/gecko-dev/js/src/jit-test/tests/ion/bu
g901086.js
Illegal instruction (core dumped)

Comment 1

[hev [:hev]]

Attachment: 0001-IonMonkey-MIPS32-Fix-ion-bug901086.js-failed.patch

Fix numStackValues conffict with jitcode. We can't add register bit before real release it.

Comment 2

[Nicolas B. Pierron [:nbp]]

Comment on attachment 8671732 [details] [diff] [review]
0001-IonMonkey-MIPS32-Fix-ion-bug901086.js-failed.patch

Review of attachment 8671732 [details] [diff] [review]:
-----------------------------------------------------------------

I guess the problem was related to the reuse of the scratch and numStackValues registers.
To be honest, I would prefer if the following calls would use regs.takeAny() instead of reusing the same named registers as we should expect a call to clobber most of the registers.

Comment 3

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/0e5643287c70

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/0e5643287c70