1885475 - [loongarch] Failed to dumpValue for floating-point value

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect, P1)

Description

[Zhao Jiazhong]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:125.0) Gecko/20100101 Firefox/125.0

Steps to reproduce:

Build js shell with --enable-debug or --enable-jitspew on loongarch64 platform, not the simulator in SpiderMonkey.

Then run: ./mach run -e 'dumpValue(1.1)'

Actual results:

JS shell crash before print all output.

Expected results:

Get the following output, and exit normally.

{
"type": "double",
"value": "1.1",
"private": "0x3ff199999999999a"
}

Comment 1

[Zhao Jiazhong]

The function and this issue seem to be introduced in bug 1783397.

Comment 2

[BugBot [:suhaib / :marco/ :calixte]]

:arai, since you are the author of the regressor, bug 1783397, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Comment 3

[Zhao Jiazhong]

The problem is there are two JSONPrinter::formatProperty:

void JSONPrinter::formatProperty(const char* name, const char* format, ...) { ... }

void JSONPrinter::formatProperty(const char* name, const char* format, va_list ap) { ... }

And va_list type is actually void * on LoongArch, and I guess RISC-V and MIPS, some codes may call into the unwanted one.

Comment 4

[Tooru Fujisawa [:arai]]

Attachment: Bug 1885475 - Use different method name for ... and va_list. r?mgaudet!

Comment 5

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1783397

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/8c3d0c204186
Use different method name for ... and va_list. r=mgaudet

Comment 7

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/8c3d0c204186