[Buildroot] Recurrent Blackfin build failure on icu

[Buildroot] Recurrent Blackfin build failure on icu

[Thomas Petazzoni]

Hello Sonic,

On our autobuilders, a number of Blackfin builds end up failing with a
problem on the icu package, which otherwise seems to build fine on
other architectures.

See
http://autobuild.buildroot.org/results/0f9/0f953610792aa75f9bb1f9218bcac27b2786e3f8/build-end.log
for one occurrence of this problem. It would be nice if you could have
a look and figure out a way of solving this build problem.

Generally speaking, would you be interested in receiving a daily e-mail
that contains the list of build failures that occurred on the previous
day, and that are related to the Blackfin architecture? Supporting an
architecture in Buildroot also means to have this architecture in a
good shape in terms of autobuilders testing.

Thanks a lot!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Recurrent Blackfin build failure on icu

[Sonic Zhang]

Hi Thomas,

For historic reasons, Blackfin GNU toolchain generate assembly symbols
with the prefix '_'. For example: _uconvmsg_dat and _icudt48_dat .
But, the pkgdata tool in icu doesn't generate assembly file
uconvmsg_dat.s and icudt48l_dat.s in this way when Blackfin GNU
toolchain is used. What do you think is a better way to fix it?

/buildroot/output/build/host-icu-4.8.1.1/source/bin/pkgdata -p
uconvmsg -O pkgdata.inc -m static -s uconvmsg -d uconvmsg -T uconvmsg
uconvmsg/uconvmsg.lst


.globl uconvmsg_dat
        .section .note.GNU-stack,"",%progbits
        .section .rodata
        .align 8
        .type uconvmsg_dat,%object
uconvmsg_dat:
...


In addition, I am fine to receive the test log by this email.

Regards,

Sonic


On Thu, May 30, 2013 at 3:35 AM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Hello Sonic,
>
> On our autobuilders, a number of Blackfin builds end up failing with a
> problem on the icu package, which otherwise seems to build fine on
> other architectures.
>
> See
> http://autobuild.buildroot.org/results/0f9/0f953610792aa75f9bb1f9218bcac27b2786e3f8/build-end.log
> for one occurrence of this problem. It would be nice if you could have
> a look and figure out a way of solving this build problem.
>
> Generally speaking, would you be interested in receiving a daily e-mail
> that contains the list of build failures that occurred on the previous
> day, and that are related to the Blackfin architecture? Supporting an
> architecture in Buildroot also means to have this architecture in a
> good shape in terms of autobuilders testing.
>
> Thanks a lot!
>
> Thomas
> --
> Thomas Petazzoni, Free Electrons
> Kernel, drivers, real-time and embedded Linux
> development, consulting, training and support.
> http://free-electrons.com

[Buildroot] Recurrent Blackfin build failure on icu

[Thomas Petazzoni]

Dear Sonic Zhang,

On Mon, 3 Jun 2013 17:26:08 +0800, Sonic Zhang wrote:

> For historic reasons, Blackfin GNU toolchain generate assembly symbols
> with the prefix '_'. For example: _uconvmsg_dat and _icudt48_dat .
> But, the pkgdata tool in icu doesn't generate assembly file
> uconvmsg_dat.s and icudt48l_dat.s in this way when Blackfin GNU
> toolchain is used. What do you think is a better way to fix it?

Can this specific behavior of the Blackfin toolchain be detected? Is
there a typical autotools test that would allow to test this?

Adding Mike in Cc, since I'm pretty sure he has some ideas about this.

> /buildroot/output/build/host-icu-4.8.1.1/source/bin/pkgdata -p
> uconvmsg -O pkgdata.inc -m static -s uconvmsg -d uconvmsg -T uconvmsg
> uconvmsg/uconvmsg.lst
> 
> 
> .globl uconvmsg_dat
>         .section .note.GNU-stack,"",%progbits
>         .section .rodata
>         .align 8
>         .type uconvmsg_dat,%object
> uconvmsg_dat:
> ...
> 
> 
> In addition, I am fine to receive the test log by this email.

Ok, I'll work on this.

Thanks,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Recurrent Blackfin build failure on icu

[Sonic Zhang]

Hi Thomas,

On Mon, Jun 3, 2013 at 5:38 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Dear Sonic Zhang,
>
> On Mon, 3 Jun 2013 17:26:08 +0800, Sonic Zhang wrote:
>
>> For historic reasons, Blackfin GNU toolchain generate assembly symbols
>> with the prefix '_'. For example: _uconvmsg_dat and _icudt48_dat .
>> But, the pkgdata tool in icu doesn't generate assembly file
>> uconvmsg_dat.s and icudt48l_dat.s in this way when Blackfin GNU
>> toolchain is used. What do you think is a better way to fix it?
>
> Can this specific behavior of the Blackfin toolchain be detected? Is
> there a typical autotools test that would allow to test this?
>

Maybe you can check the symbol name after compiling a C file into an
assembly file?

Regards,

Sonic

[Buildroot] Recurrent Blackfin build failure on icu

[Thomas Petazzoni]

Dear Sonic Zhang,

On Thu, 6 Jun 2013 10:32:27 +0800, Sonic Zhang wrote:

> > Can this specific behavior of the Blackfin toolchain be detected? Is
> > there a typical autotools test that would allow to test this?
> >
> 
> Maybe you can check the symbol name after compiling a C file into an
> assembly file?

Good idea, thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Recurrent Blackfin build failure on icu

[Mike Frysinger]

On Wednesday 05 June 2013 22:32:27 Sonic Zhang wrote:
> On Mon, Jun 3, 2013 at 5:38 PM, Thomas Petazzoni wrote:
> > On Mon, 3 Jun 2013 17:26:08 +0800, Sonic Zhang wrote:
> >> For historic reasons, Blackfin GNU toolchain generate assembly symbols
> >> with the prefix '_'. For example: _uconvmsg_dat and _icudt48_dat .
> >> But, the pkgdata tool in icu doesn't generate assembly file
> >> uconvmsg_dat.s and icudt48l_dat.s in this way when Blackfin GNU
> >> toolchain is used. What do you think is a better way to fix it?
> > 
> > Can this specific behavior of the Blackfin toolchain be detected? Is
> > there a typical autotools test that would allow to test this?
> 
> Maybe you can check the symbol name after compiling a C file into an
> assembly file?

gcc provides __USER_LABEL_PREFIX__ specifically for this.  but you'd have to 
create a .S file and run it through the preprocessor to do that ...
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20130620/c44662b3/attachment.asc>

[Buildroot] Recurrent Blackfin build failure on icu

[Thomas Petazzoni]

Dear Mike Frysinger,

On Thu, 20 Jun 2013 19:21:48 -0400, Mike Frysinger wrote:

> > Maybe you can check the symbol name after compiling a C file into an
> > assembly file?
> 
> gcc provides __USER_LABEL_PREFIX__ specifically for this.  but you'd have to 
> create a .S file and run it through the preprocessor to do that ...

So you mean that __USER_LABEL_PREFIX__ is "_" on Blackfin and empty on
other architectures, so that assembly code should look like:

	__USER_LABEL_PREFIX__ ## symbol_foobar

and run through the preprocessor, to work properly across arches?

Thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com