Bug 124370

Summary:

Fix failed to build for Linux/MIPS64EL

Product:

WebKit

Reporter:

YunQiang Su <wzssyqa>

Component:

JavaScriptCore

Assignee:

Nobody <webkit-unassigned>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

benjamin, berto, cmarcelo, commit-queue, ggaren

Priority:

Version:

528+ (Nightly build)

Hardware:

All

OS:

Linux

Attachments:

Description	Flags
patch for build webkit on Linux/MIPS64EL	none
patch merged into qtwebkit.	none
Patch	none
Patch	none
don't use __mips64 for N64 determination.	darin: review+

Description YunQiang Su 2013-11-14 10:14:46 PST

There is no process for 64bit mips processors,
so for now, it will use JSVAULE32_64 instead of JSCVALUE64

With the attached patch, I can build webkitgtk (2.2.0) and
qtwebkit 2.2.1 on Debian Sid.

Comment 1 YunQiang Su 2013-11-14 10:15:46 PST

Created attachment 216957 [details]
patch for build webkit on Linux/MIPS64EL

Comment 2 YunQiang Su 2014-07-10 03:02:34 PDT

This patch has been merged into qtwebkit, and test well.

I also test it for webkitgtk.

Please consider merge it.

Comment 3 YunQiang Su 2014-07-10 03:04:51 PDT

Created attachment 234697 [details]
patch merged into qtwebkit.

Comment 4 Alberto Garcia 2014-07-10 05:40:25 PDT

Created attachment 234705 [details]
Patch

The patch needs a ChangeLog entry, I just added one.

Anyway I think the addition of new architectures must be discussed first in webkit-dev, am I right?

Comment 5 Geoffrey Garen 2014-07-11 14:25:18 PDT

Using JSValue32_64 on a 64bit architecture is pretty weak sauce. (And does it work at all, given that pointers are 64bit?)

Is there anyone with enough involvement in the project, who cares about this architecture, who can make it work with JSValue64?

Comment 6 YunQiang Su 2014-07-11 20:01:06 PDT

In this patch, we use JSValue64 not JSValue32_64.

Comment 7 YunQiang Su 2014-10-29 23:46:47 PDT

Any progress of this bug?

Comment 8 Alberto Garcia 2014-11-02 11:21:08 PST

This has been working in Debian for a while now, can anyone review this?

Comment 9 Alberto Garcia 2015-01-18 11:57:20 PST

Created attachment 244865 [details]
Patch

I've just rebased the patch. Can someone please review?

Comment 10 YunQiang Su 2015-01-19 01:12:15 PST

Sorry a problem discovered

#if defined(__mips64)

should be

defined(_MIPS_SIM_ABI64) && (_MIPS_SIM == _MIPS_SIM_ABI64)

as, __mips64 is also defined for N32 abi.

Comment 11 YunQiang Su 2015-01-19 01:15:23 PST

Created attachment 244881 [details]
don't use __mips64 for N64 determination.

Comment 12 Alberto Garcia 2015-01-19 01:19:27 PST

Comment on attachment 244881 [details]
don't use __mips64 for N64 determination.

Ok, thanks

Comment 13 Darin Adler 2015-01-20 09:27:12 PST

Comment on attachment 244881 [details]
don't use __mips64 for N64 determination.

View in context: https://bugs.webkit.org/attachment.cgi?id=244881&action=review

I’m not a platform expert, but this patch looks OK to me.

> Source/WTF/wtf/Platform.h:85
> +#if (defined(mips) || defined(__mips__) || defined(MIPS) || defined(_MIPS_) \
> +    || defined(__mips64))

This line break just makes things hard to read. We are not trying to fit in 80 columns. Please consider putting this all on one line.

Comment 14 Alberto Garcia 2015-01-20 11:10:58 PST

(In reply to comment #13)
> > +#if (defined(mips) || defined(__mips__) || defined(MIPS) || defined(_MIPS_) \
> > +    || defined(__mips64))
>
> This line break just makes things hard to read. We are not trying to
> fit in 80 columns. Please consider putting this all on one line.

Sure, I'll do that.

Thanks!

Comment 15 Alberto Garcia 2015-01-20 11:12:29 PST

Committed r178725: <http://trac.webkit.org/changeset/178725>