[Buildroot] How to bypass fix-rpath for a particular package?

[Buildroot] How to bypass fix-rpath for a particular package?

[Tudor Holton]

Hi all,

I have a br_external package that always lands corrupt on the image and 
segfaults. I tried a number of things and eventually discovered that it 
works fine if I graft the output from build directly into the final 
image.? From this I narrowed it down to something that happens during 
fix-rpath, which looks simple, but all my checks show that I'm not 
linking with anything other than -rpath=$ORIGIN so I'm at a loss as to 
the root cause.

This may not be the best solution, but is there a standard way to tell 
fix-rpath not to patchelf my package and just copy it verbatim to the image?

Cheers,
Tudor.

[Buildroot] How to bypass fix-rpath for a particular package?

[Bryce Ferguson]

Hello Tudor,

When you say that the file is corrupted on the image, what does that mean
exactly?  Have you performed a readelf on the binary to determine what
fields may have been messed up?  What is your target architecture?

Thanks,

-Bryce

On Mon, Jul 16, 2018 at 10:53 PM, Tudor Holton <tudor@smartguide.com.au>
wrote:

> Hi all,
>
> I have a br_external package that always lands corrupt on the image and
> segfaults. I tried a number of things and eventually discovered that it
> works fine if I graft the output from build directly into the final image.
> From this I narrowed it down to something that happens during fix-rpath,
> which looks simple, but all my checks show that I'm not linking with
> anything other than -rpath=$ORIGIN so I'm at a loss as to the root cause.
>
> This may not be the best solution, but is there a standard way to tell
> fix-rpath not to patchelf my package and just copy it verbatim to the image?
>
> Cheers,
> Tudor.
>
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot
>



-- 


*Bryce Ferguson*
Software Engineer
Airborne Information Solutions \ RC Linux Platform

400 Collins Rd NE Cedar Rapids, Iowa 52498

Bryce.Ferguson at rockwellcollins.com <bryce.ferguson@rockwellcollins.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20180717/42bc25bd/attachment.html>

[Buildroot] How to bypass fix-rpath for a particular package?

[Tudor Holton]

Hi Bryce,

It appears that the symbols referencing the library names have been 
bitshifted.? At best, objdump -x gives correct references, ldd 
segfaults.? At worst, the objdump -x contains structured gibberish.? It 
doesn't happen on all files, though, but at this point I'm unable to 
discern a pattern.

I'm guessing that patchelf has a false positive, but I'm no expert in 
this area.? I'm not used to ldd segfaulting so I have only been learning 
this stuff since I discovered the behaviour.? It came as a complete 
surprise to me when I noticed the md5sum difference and grafted directly 
and it worked.

Cheers,
Tudor.

On 17/07/18 23:55, Bryce Ferguson wrote:
> Hello Tudor,
>
> When you say that the file is corrupted on the image, what does that 
> mean exactly?? Have you performed a readelf on the binary to determine 
> what fields may have been messed up? What is your target architecture?
>
> Thanks,
>
> -Bryce
>
> On Mon, Jul 16, 2018 at 10:53 PM, Tudor Holton 
> <tudor at smartguide.com.au <mailto:tudor@smartguide.com.au>> wrote:
>
>     Hi all,
>
>     I have a br_external package that always lands corrupt on the
>     image and segfaults. I tried a number of things and eventually
>     discovered that it works fine if I graft the output from build
>     directly into the final image.? From this I narrowed it down to
>     something that happens during fix-rpath, which looks simple, but
>     all my checks show that I'm not linking with anything other than
>     -rpath=$ORIGIN so I'm at a loss as to the root cause.
>
>     This may not be the best solution, but is there a standard way to
>     tell fix-rpath not to patchelf my package and just copy it
>     verbatim to the image?
>
>     Cheers,
>     Tudor.
>
>     _______________________________________________
>     buildroot mailing list
>     buildroot at busybox.net <mailto:buildroot@busybox.net>
>     http://lists.busybox.net/mailman/listinfo/buildroot
>     <http://lists.busybox.net/mailman/listinfo/buildroot>
>
>
>
>
> -- 
> *
> *
> *
> *
> *Bryce Ferguson*
> Software Engineer
> Airborne Information Solutions \ RC Linux Platform
>
> 400 Collins Rd NE Cedar Rapids, Iowa 52498
>
> Bryce.Ferguson at rockwellcollins.com 
> <mailto:bryce.ferguson@rockwellcollins.com>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20180718/e31e7a5c/attachment.html>

[Buildroot] How to bypass fix-rpath for a particular package?

[Bryce Ferguson]

Tudor,

Please provide your target architecture and the version of buildroot you
are using.

Thanks!

On Tue, Jul 17, 2018 at 7:05 PM, Tudor Holton <tudor@smartguide.com.au>
wrote:

> Hi Bryce,
>
> It appears that the symbols referencing the library names have been
> bitshifted.  At best, objdump -x gives correct references, ldd segfaults.
> At worst, the objdump -x contains structured gibberish.  It doesn't happen
> on all files, though, but at this point I'm unable to discern a pattern.
>
> I'm guessing that patchelf has a false positive, but I'm no expert in this
> area.  I'm not used to ldd segfaulting so I have only been learning this
> stuff since I discovered the behaviour.  It came as a complete surprise to
> me when I noticed the md5sum difference and grafted directly and it worked.
>
> Cheers,
> Tudor.
>
>
> On 17/07/18 23:55, Bryce Ferguson wrote:
>
> Hello Tudor,
>
> When you say that the file is corrupted on the image, what does that mean
> exactly?  Have you performed a readelf on the binary to determine what
> fields may have been messed up?  What is your target architecture?
>
> Thanks,
>
> -Bryce
>
> On Mon, Jul 16, 2018 at 10:53 PM, Tudor Holton <tudor@smartguide.com.au>
> wrote:
>
>> Hi all,
>>
>> I have a br_external package that always lands corrupt on the image and
>> segfaults. I tried a number of things and eventually discovered that it
>> works fine if I graft the output from build directly into the final image.
>> From this I narrowed it down to something that happens during fix-rpath,
>> which looks simple, but all my checks show that I'm not linking with
>> anything other than -rpath=$ORIGIN so I'm at a loss as to the root cause.
>>
>> This may not be the best solution, but is there a standard way to tell
>> fix-rpath not to patchelf my package and just copy it verbatim to the image?
>>
>> Cheers,
>> Tudor.
>>
>> _______________________________________________
>> buildroot mailing list
>> buildroot at busybox.net
>> http://lists.busybox.net/mailman/listinfo/buildroot
>>
>
>
>
> --
>
>
> *Bryce Ferguson*
> Software Engineer
> Airborne Information Solutions \ RC Linux Platform
>
> 400 Collins Rd NE Cedar Rapids, Iowa 52498
> <https://maps.google.com/?q=400+Collins+Rd+NE+Cedar+Rapids,+Iowa+52498&entry=gmail&source=g>
>
> Bryce.Ferguson at rockwellcollins.com <bryce.ferguson@rockwellcollins.com>
>
>
>


-- 


*Bryce Ferguson*
Software Engineer
Airborne Information Solutions \ RC Linux Platform

400 Collins Rd NE Cedar Rapids, Iowa 52498

Bryce.Ferguson at rockwellcollins.com <bryce.ferguson@rockwellcollins.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20180718/4396cfa2/attachment.html>

[Buildroot] How to bypass fix-rpath for a particular package?

[Tudor Holton]

Hi Bryce,

My apologies.? I realised I hadn't provided that detail only as I was 
walking home last night.

I'm currently building a Intel x86_64 uclibc (with getc/putc macros 
disabled) buildroot Image on an Intel x86_64 glibc (Ubuntu) docker 
container on an Intel x86_64 glibc (Ubuntu) host.

Cheers,
Tudor.

On 19/07/18 00:20, Bryce Ferguson wrote:
> Tudor,
>
> Please provide your target architecture and the version of buildroot 
> you are using.
>
> Thanks!
>
> On Tue, Jul 17, 2018 at 7:05 PM, Tudor Holton <tudor@smartguide.com.au 
> <mailto:tudor@smartguide.com.au>> wrote:
>
>     Hi Bryce,
>
>     It appears that the symbols referencing the library names have
>     been bitshifted.? At best, objdump -x gives correct references,
>     ldd segfaults.? At worst, the objdump -x contains structured
>     gibberish.? It doesn't happen on all files, though, but at this
>     point I'm unable to discern a pattern.
>
>     I'm guessing that patchelf has a false positive, but I'm no expert
>     in this area.? I'm not used to ldd segfaulting so I have only been
>     learning this stuff since I discovered the behaviour.? It came as
>     a complete surprise to me when I noticed the md5sum difference and
>     grafted directly and it worked.
>
>     Cheers,
>     Tudor.
>
>
>     On 17/07/18 23:55, Bryce Ferguson wrote:
>>     Hello Tudor,
>>
>>     When you say that the file is corrupted on the image, what does
>>     that mean exactly?? Have you performed a readelf on the binary to
>>     determine what fields may have been messed up?? What is your
>>     target architecture?
>>
>>     Thanks,
>>
>>     -Bryce
>>
>>     On Mon, Jul 16, 2018 at 10:53 PM, Tudor Holton
>>     <tudor at smartguide.com.au <mailto:tudor@smartguide.com.au>> wrote:
>>
>>         Hi all,
>>
>>         I have a br_external package that always lands corrupt on the
>>         image and segfaults. I tried a number of things and
>>         eventually discovered that it works fine if I graft the
>>         output from build directly into the final image.? From this I
>>         narrowed it down to something that happens during fix-rpath,
>>         which looks simple, but all my checks show that I'm not
>>         linking with anything other than -rpath=$ORIGIN so I'm at a
>>         loss as to the root cause.
>>
>>         This may not be the best solution, but is there a standard
>>         way to tell fix-rpath not to patchelf my package and just
>>         copy it verbatim to the image?
>>
>>         Cheers,
>>         Tudor.
>>
>>         _______________________________________________
>>         buildroot mailing list
>>         buildroot at busybox.net <mailto:buildroot@busybox.net>
>>         http://lists.busybox.net/mailman/listinfo/buildroot
>>         <http://lists.busybox.net/mailman/listinfo/buildroot>
>>
>>
>>
>>
>>     -- 
>>     *
>>     *
>>     *
>>     *
>>     *Bryce Ferguson*
>>     Software Engineer
>>     Airborne Information Solutions \ RC Linux Platform
>>
>>     400 Collins Rd NE Cedar Rapids, Iowa 52498
>>     <https://maps.google.com/?q=400+Collins+Rd+NE+Cedar+Rapids,+Iowa+52498&entry=gmail&source=g>
>>
>>     Bryce.Ferguson at rockwellcollins.com
>>     <mailto:bryce.ferguson@rockwellcollins.com>
>>
>
>
>
>
> -- 
> *
> *
> *
> *
> *Bryce Ferguson*
> Software Engineer
> Airborne Information Solutions \ RC Linux Platform
>
> 400 Collins Rd NE Cedar Rapids, Iowa 52498
>
> Bryce.Ferguson at rockwellcollins.com 
> <mailto:bryce.ferguson@rockwellcollins.com>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20180719/893a2da0/attachment.html>