[meta-browser] firefox not working for me (only)

[meta-browser] firefox not working for me (only)

[Andreas Müller]

Hi Gary,

Turned off-topic so I opened a new thread. For those who missed the
old thread [1] in short:

for me

* firefox fails with illeagal ARM instruction
* chromium fails during start - details I can't reproduce currently


On Wed, Jun 17, 2015 at 3:32 PM, Gary Thomas <gary@mlbassoc.com> wrote:
>> Hmm - I was wondering why nobody took care.
>>
>> All my plattforms fail:
>> * i.MX6 base
>> * gumstix
>> * raspberry Pi2
>>
>> I'll put this browser issue on my TODO currently I have no further
>> time to take care.
>
>
> BTW, I just tested both chromium and firefox on the i.MX6Q and RPi2
> and they worked just fine.  Built from meta-browser:0cc0419a
>
Thanks for information. Let's focus on firefox's illegal ARM
instruction crasher:

* Last time I felt the only one facing a specific error it was caused
by angstrom enabling thumb(2) for ARM (long story @meta-fsl-arm).
Gary: I assume, you don't have thumb enabled - right?
* firefox doesn't crash with "--safe-mode"

Looking into meta-browser I saw that firefox pins ARM_INSTRUCTION_SET
= "arm" but firefox-l10n does not. When having some time left over
(=never) I will test with thumb disabled for firefox-l10n and
chromium.

Andreas

[1] http://lists.openembedded.org/pipermail/openembedded-core/2015-June/106229.html

Re: [meta-browser] firefox not working for me (only)

[Gary Thomas]

On 2015-06-17 08:36, Andreas Müller wrote:
> Hi Gary,
>
> Turned off-topic so I opened a new thread. For those who missed the
> old thread [1] in short:
>
> for me
>
> * firefox fails with illeagal ARM instruction
> * chromium fails during start - details I can't reproduce currently
>
>
> On Wed, Jun 17, 2015 at 3:32 PM, Gary Thomas <gary@mlbassoc.com> wrote:
>>> Hmm - I was wondering why nobody took care.
>>>
>>> All my plattforms fail:
>>> * i.MX6 base
>>> * gumstix
>>> * raspberry Pi2
>>>
>>> I'll put this browser issue on my TODO currently I have no further
>>> time to take care.
>>
>>
>> BTW, I just tested both chromium and firefox on the i.MX6Q and RPi2
>> and they worked just fine.  Built from meta-browser:0cc0419a
>>
> Thanks for information. Let's focus on firefox's illegal ARM
> instruction crasher:
>
> * Last time I felt the only one facing a specific error it was caused
> by angstrom enabling thumb(2) for ARM (long story @meta-fsl-arm).
> Gary: I assume, you don't have thumb enabled - right?
> * firefox doesn't crash with "--safe-mode"

I am not using thumb - off the shelf configurations for RaspberryPi2 and
BoundaryDevices nitrogen6x built with $DISTRO=poky are the ones I've tested.

>
> Looking into meta-browser I saw that firefox pins ARM_INSTRUCTION_SET
> = "arm" but firefox-l10n does not. When having some time left over
> (=never) I will test with thumb disabled for firefox-l10n and
> chromium.
>
> Andreas
>
> [1] http://lists.openembedded.org/pipermail/openembedded-core/2015-June/106229.html
>

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: [meta-browser] firefox not working for me (only)

[Max Krummenacher]

> > * firefox fails with illeagal ARM instruction

That sounds familiar, I thought then that it was specific to our setup
with the binary Tegra X driver and the armv7a without neon setup, that's
why I did not upstream my patches.

> > On Wed, Jun 17, 2015 at 3:32 PM, Gary Thomas <gary@mlbassoc.com> wrote:
> >>> Hmm - I was wondering why nobody took care.
> >>>
> >>> All my plattforms fail:
> >>> * i.MX6 base
> >>> * gumstix
> >>> * raspberry Pi2
> >>>
> >>> I'll put this browser issue on my TODO currently I have no further
> >>> time to take care.
> >>
> >>
> >> BTW, I just tested both chromium and firefox on the i.MX6Q and RPi2
> >> and they worked just fine.  Built from meta-browser:0cc0419a
> >>
> > Thanks for information. Let's focus on firefox's illegal ARM
> > instruction crasher:
> >
> > * Last time I felt the only one facing a specific error it was caused
> > by angstrom enabling thumb(2) for ARM (long story @meta-fsl-arm).
> > Gary: I assume, you don't have thumb enabled - right?
> > * firefox doesn't crash with "--safe-mode"
> 
> I am not using thumb - off the shelf configurations for RaspberryPi2 and
> BoundaryDevices nitrogen6x built with $DISTRO=poky are the ones I've tested.
> 
Actually I use Angstrom with the instruction set defaulting to ARM.

> >
> > Looking into meta-browser I saw that firefox pins ARM_INSTRUCTION_SET
> > = "arm" but firefox-l10n does not. When having some time left over
> > (=never) I will test with thumb disabled for firefox-l10n and
> > chromium.
> >
> > Andreas

Have a look at
http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-mozilla/firefox?h=V2.4-next

alignment.patch,
Prevent-SIGILL-crashes-on-ARMv6-builds-built-with-the-NDK_10.0esr.patch .

If usefull I could send a patch for meta-browser.

Regards,
Max

Re: [meta-browser] firefox not working for me (only)

[Otavio Salvador]

On Wed, Jun 17, 2015 at 5:18 PM, Max Krummenacher <max.oss.09@gmail.com> wrote:
>> > * firefox fails with illeagal ARM instruction
>
> That sounds familiar, I thought then that it was specific to our setup
> with the binary Tegra X driver and the armv7a without neon setup, that's
> why I did not upstream my patches.
>
>> > On Wed, Jun 17, 2015 at 3:32 PM, Gary Thomas <gary@mlbassoc.com> wrote:
>> >>> Hmm - I was wondering why nobody took care.
>> >>>
>> >>> All my plattforms fail:
>> >>> * i.MX6 base
>> >>> * gumstix
>> >>> * raspberry Pi2
>> >>>
>> >>> I'll put this browser issue on my TODO currently I have no further
>> >>> time to take care.
>> >>
>> >>
>> >> BTW, I just tested both chromium and firefox on the i.MX6Q and RPi2
>> >> and they worked just fine.  Built from meta-browser:0cc0419a
>> >>
>> > Thanks for information. Let's focus on firefox's illegal ARM
>> > instruction crasher:
>> >
>> > * Last time I felt the only one facing a specific error it was caused
>> > by angstrom enabling thumb(2) for ARM (long story @meta-fsl-arm).
>> > Gary: I assume, you don't have thumb enabled - right?
>> > * firefox doesn't crash with "--safe-mode"
>>
>> I am not using thumb - off the shelf configurations for RaspberryPi2 and
>> BoundaryDevices nitrogen6x built with $DISTRO=poky are the ones I've tested.
>>
> Actually I use Angstrom with the instruction set defaulting to ARM.
>
>> >
>> > Looking into meta-browser I saw that firefox pins ARM_INSTRUCTION_SET
>> > = "arm" but firefox-l10n does not. When having some time left over
>> > (=never) I will test with thumb disabled for firefox-l10n and
>> > chromium.
>> >
>> > Andreas
>
> Have a look at
> http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-mozilla/firefox?h=V2.4-next
>
> alignment.patch,
> Prevent-SIGILL-crashes-on-ARMv6-builds-built-with-the-NDK_10.0esr.patch .
>
> If usefull I could send a patch for meta-browser.

An upgraded version would be awesome :-)

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [meta-browser] firefox not working for me (only)

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1963 bytes --]


> On Jun 17, 2015, at 1:18 PM, Max Krummenacher <max.oss.09@gmail.com> wrote:
> 
>>> * firefox fails with illeagal ARM instruction
> 
> That sounds familiar, I thought then that it was specific to our setup
> with the binary Tegra X driver and the armv7a without neon setup, that's
> why I did not upstream my patches.
> 
>>> On Wed, Jun 17, 2015 at 3:32 PM, Gary Thomas <gary@mlbassoc.com> wrote:
>>>>> Hmm - I was wondering why nobody took care.
>>>>> 
>>>>> All my plattforms fail:
>>>>> * i.MX6 base
>>>>> * gumstix
>>>>> * raspberry Pi2
>>>>> 
>>>>> I'll put this browser issue on my TODO currently I have no further
>>>>> time to take care.
>>>> 
>>>> 
>>>> BTW, I just tested both chromium and firefox on the i.MX6Q and RPi2
>>>> and they worked just fine.  Built from meta-browser:0cc0419a
>>>> 
>>> Thanks for information. Let's focus on firefox's illegal ARM
>>> instruction crasher:
>>> 
>>> * Last time I felt the only one facing a specific error it was caused
>>> by angstrom enabling thumb(2) for ARM (long story @meta-fsl-arm).
>>> Gary: I assume, you don't have thumb enabled - right?
>>> * firefox doesn't crash with "--safe-mode"
>> 
>> I am not using thumb - off the shelf configurations for RaspberryPi2 and
>> BoundaryDevices nitrogen6x built with $DISTRO=poky are the ones I've tested.
>> 
> Actually I use Angstrom with the instruction set defaulting to ARM.
> 
>>> 
>>> Looking into meta-browser I saw that firefox pins ARM_INSTRUCTION_SET
>>> = "arm" but firefox-l10n does not. When having some time left over
>>> (=never) I will test with thumb disabled for firefox-l10n and
>>> chromium.
>>> 
>>> Andreas
> 
> Have a look at
> http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-mozilla/firefox?h=V2.4-next
> 
> alignment.patch,
> Prevent-SIGILL-crashes-on-ARMv6-builds-built-with-the-NDK_10.0esr.patch .
> 
> If usefull I could send a patch for meta-browser.

Please do so.

[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]

Re: [meta-browser] firefox not working for me (only)

[S. Lockwood-Childs]

On Sat, Aug 08, 2015 at 04:01:38PM -0700, Khem Raj wrote:
> 
> > On Jul 17, 2015, at 5:38 PM, S. Lockwood-Childs <sjl@vctlabs.com> wrote:
> > 
> > configure.ac had a runtime check for large file support that got run
> > on 64-bit architectures, and all runtime checks must go away to be
> > cross-compile safe.
> > 
> > Luckily that runtime check was extraneous, so just remove it;
> > standard AC_SYS_LARGEFILE macro should be sufficient for figuring out
> > proper settings to get large file support.
> > 
> > Also un-blacklist libimobiledevice, since this problem was the
> > reason for the blacklist.
> > 
> 
> one difference is that now it will depend on --enable-largefile secondly, what does AC_SYS_LARGEFILE
> check do in the end ? is it a link time check ?

Not sure what you mean by 'now it will depend on --enable-largefile' since 
the handling of that flag isn't being changed. I didn't add AC_SYS_LARGEFILE
macro; it was already there. The recipe still doesn't need to set that flag...
large file support should get enabled if the platform supports it.

AC_SYS_LARGEFILE does a compile-time check that off_t is large as expected,
and if not it tries some defines (such as _FILE_OFFSET_BITS) to see if they
can make off_t check pass. In the 64-bit Linux case, the first check passes,
since no tinkering with extra defines is needed, but for 32-bit case 
_FILE_OFFSET_BITS will get defined.