527363 - JS-ctypes fail to compile on Maemo ARM.

Status: RESOLVED FIXED

(Core :: js-ctypes, defect)

Description

[Oleg Romashin (:romaxa)]

Attachment: Patch from bug URL, arm softfp

gcc -DHAVE_CONFIG_H -I. -Ijs/ctypes/libffi -I. -Ijs/ctypes/libffi/include -Iinclude -Ijs/ctypes/libffi/src -Wall -g -fexceptions -g -O2 -MT src/arm/ffi.lo -MD -MP -MF src/arm/.deps/ffi.Tpo -c js/ctypes/libffi/src/arm/ffi.c  -fPIC -DPIC -o src/arm/ffi.o
js/ctypes/libffi/src/arm/sysv.S: Assembler messages:
js/ctypes/libffi/src/arm/sysv.S:203: Error: selected processor does not support `stfeqs f0,[r2]'
js/ctypes/libffi/src/arm/sysv.S:208: Error: selected processor does not support `stfeqd f0,[r2]'
js/ctypes/libffi/src/arm/sysv.S:283: Error: selected processor does not support `ldfs f0,[sp]'
js/ctypes/libffi/src/arm/sysv.S:286: Error: selected processor does not support `ldfd f0,[sp]'
js/ctypes/libffi/src/arm/sysv.S:289: Error: selected processor does not support `ldfd f0,[sp]'

Comment 1

[[e-mail]]

Comment on attachment 411080 [details] [diff] [review]
Patch from bug URL, arm softfp

>diff -r a2205b22d77c configure.in
>--- a/configure.in	Sat Nov 07 20:28:54 2009 +0200
>+++ b/configure.in	Sun Nov 08 22:52:00 2009 +0200
>@@ -8165,8 +8165,6 @@ dnl Build jsctypes on the platforms we c
> dnl
> AC_SUBST(BUILD_CTYPES)
> case "$OS_TEST" in
>-arm*)
>-  ;;
> *)
>   BUILD_CTYPES=1
>   AC_DEFINE(BUILD_CTYPES)

This will enable ctypes for wince/arm as well, which we don't want. You'll have to call out linux/arm specifically.

Does this pass ctypes unit tests locally and on tryserver?

Thanks for the help here, this is great to see!

Comment 3

[Oleg Romashin (:romaxa)]

> 
> This will enable ctypes for wince/arm as well, which we don't want. You'll have
> to call out linux/arm specifically.

Yep, I see build failure on wince and winmo

What is the best define for it? should I test it for linux only in arm case?


> 
> Does this pass ctypes unit tests locally and on tryserver?

I try it locally on N900:
https://developer.mozilla.org/en/JavaScript_code_modules/ctypes.jsm

Nokia-N900-46-2:/usr/lib/microb-engine# ./run-mozilla.sh ./xpcshell 
defined DISPLAY=:0.0
js> Components.utils.import("resource://gre/modules/ctypes.jsm");
[object BackstagePass]
js> var libc = ctypes.open("libc.so.6"); 
js> var puts = libc.declare("puts",ctypes.default_abi,ctypes.int32_t,ctypes.string);
js> var ret = puts("Hello World from js-ctypes!");
Hello World from js-ctypes!
js> 

It seems to works fine.

Comment 4

[Oleg Romashin (:romaxa)]

Attachment: TryServer arm_ctypes.v2

Comment 5

[[e-mail]]

(In reply to comment #3)
> What is the best define for it? should I test it for linux only in arm case?

Instead of the case statement, you could just disable ctypes only for OS_TARGET == WINCE and filter OS_TEST for arm.

> I try it locally on N900:

We have unit tests for it - "cd $OBJDIR/js/ctypes/tests && make xpcshell-tests" will run them.

Comment 6

[Oleg Romashin (:romaxa)]

Attachment: Fixed comments

Try server results for updated patch:
https://build.mozilla.org/tryserver-builds/[e-mail]-arm_ctypes.v3/

Comment 7

[[e-mail]]

Comment on attachment 411392 [details] [diff] [review]
Fixed comments

Awesome. Let's get this on trunk and then look at a branch landing. I'll check with Stuart if we need to do anything special to get this into Fennec (and if, indeed, they want it).

Comment 8

[Oleg Romashin (:romaxa)]

Pushed in:
http://hg.mozilla.org/mozilla-central/rev/c8f2a90879fb

Comment 9

[Mark Finkle (:mfinkle) (use needinfo?)]

Comment on attachment 411392 [details] [diff] [review]
Fixed comments

I think this would be a nice addition to Fennec's binary extension story.

Comment 10

[Fabrice Desré]

Any chance to see this patch land on 1.9.2 ? I have a use case with gstreamer that works well in fennec on trunk on a n810 (http://blog.desre.org/post/2009/11/19/Using-GStreamer-with-js-ctypes), and it would be great to use it with fennec 1.0

Comment 11

[Benjamin Smedberg]

Comment on attachment 411392 [details] [diff] [review]
Fixed comments

Was this patch landed upstream in the ctypes repo?

Comment 12

[Mark Finkle (:mfinkle) (use needinfo?)]

(In reply to comment #11)
> (From update of attachment 411392 [details] [diff] [review])
> Was this patch landed upstream in the ctypes repo?

I don't think it was. Do we have a process for that?

Comment 13

[[e-mail]]

(In reply to comment #12)
> I don't think it was. Do we have a process for that?

It was; see gcc bug in the URL. (If it hasn't already been synced to libffi proper, they'll pick it up eventually.)

Comment 14

[Mark Finkle (:mfinkle) (use needinfo?)]

pushed to m-192:
http://hg.mozilla.org/releases/mozilla-1.9.2/rev/5edb87f17819

Comment 15

[[e-mail]]

Pinged Anthony about getting this into libffi 3.0.9.