1163346 - Add NSPR support for FreeBSD mips/mips64

Status: RESOLVED FIXED

(NSPR :: NSPR, defect, P2)

Description

[Jan Beich]

See Also downstream bug for details.

Comment 1

[Jan Beich]

Attachment: v1

Comment 2

[Wan-Teh Chang]

Comment on attachment 8603764 [details] [diff] [review]
v1

Review of attachment 8603764 [details] [diff] [review]:
-----------------------------------------------------------------

r=wtc. Thanks for the patch. I have a question about the endianness.

Patch checked in: https://hg.mozilla.org/projects/nspr/rev/955bfb84eb9d

::: pr/include/md/_freebsd.cfg
@@ +394,5 @@
>  
> +#elif defined(__mips__)
> +
> +#undef  IS_LITTLE_ENDIAN
> +#define IS_BIG_ENDIAN    1

These days MIPS is often configured in little-endian mode
to make it easier to port code that doesn't handle byte
order correctly. We should test some compiler predefined
macro to determine the endianness. If we hardcode the
endianness, we probably should use little endian for MIPS
today.

Could you please look into this? Thanks.

Comment 3

[Jan Beich]

Attachment: followup

Another issue was __mips__ being defined for both mips and mips64.

(In reply to Wan-Teh Chang from comment #2)
> We should test some compiler predefined macro to determine the endianness.

Clang uses _MIPSEL and _MIPSEB which may break on _linux.cfg as well.

> If we hardcode the endianness, we probably should use little endian for MIPS today.

Maintenance-wise spliting out arch-specific constants into OS-agnostic .cfg file would make more sense.

Comment 4

[Wan-Teh Chang]

Comment on attachment 8604249 [details] [diff] [review]
followup

r=wtc. It is unfortunate that clang and gcc don't use the same
macros for endianness detection.

Patch checked in: https://hg.mozilla.org/projects/nspr/rev/2f45c6fd036e