Re: [PATCH] prelink: Added prelink-20100106, which works on ARM.

[Mark Hatle <mark.hatle@windriver.com>]

Have you looked at:

http://git.yoctoproject.org/cgit/cgit.cgi/prelink-cross/

As far as I'm aware, this version works properly for all of the expected
targets, including ARM.

If it doesn't work, someone should submit a patch to the maintainer (me) and
we'll get things resolved.

The Yocto prelinker contains both local (on-target) as well as cross-prelinking
support... (This is already integrated into oe-core)

--Mark

On 6/14/11 8:16 PM, James Limbouris wrote:
> On Tuesday, 14 June 2011 5:17 PM, Phil Blundell wrote:
>> On Tue, 2011-06-14 at 15:49 +0800, James Limbouris wrote:
>>> This version of prelink, together with a patch by Carsten Munk
>>> <carsten@maemo.org>, taken from MeeGo, fixes an ARM incompatibility
>>> with 20090925.
>>>
>>> Signed-off-by: James Limbouris <james@digitalmatter.com.au>
>>> ---
>>>  .../prelink-20100106-arm-fix.patch                 |   86 ++++++++++++++++++++
>>>  recipes/prelink/prelink_20100106.bb                |   53 ++++++++++++
>>>  2 files changed, 139 insertions(+), 0 deletions(-)  create mode
>>> 100755 recipes/prelink/prelink-20100106/prelink-20100106-arm-fix.patch
>>>  create mode 100644 recipes/prelink/prelink_20100106.bb
>>>
>>> diff --git
>>> a/recipes/prelink/prelink-20100106/prelink-20100106-arm-fix.patch
>>> b/recipes/prelink/prelink-20100106/prelink-20100106-arm-fix.patch
>>> new file mode 100755
>>> index 0000000..a694541
>>> --- /dev/null
>>> +++ b/recipes/prelink/prelink-20100106/prelink-20100106-arm-fix.patch
>>> @@ -0,0 +1,86 @@
>>> +diff -ru prelink-old/src/arch-arm.c prelink/src/arch-arm.c
>>> +--- prelink-old/src/arch-arm.c	2009-06-15 07:37:50.000000000 -0400
>>> ++++ prelink/src/arch-arm.c	2010-08-17 03:35:05.000000000 -0400
>>> +@@ -832,7 +832,7 @@
>>> +   .R_COPY = R_ARM_COPY,
>>> +   .R_RELATIVE = R_ARM_RELATIVE,
>>> +   .rtype_class_valid = RTYPE_CLASS_VALID,
>>> +-  .dynamic_linker = "/lib/ld-linux.so.2",
>>> ++  .dynamic_linker = "/lib/ld-linux.so.3",
>>
>> That looks like it's going to break OABI binaries.  I'm not sure to what extent
>> we care about those anymore, but I'm not sure I would describe this as a
>> "fix" exactly.
>>
>>> +-  switch (vfork ())
>>> ++  switch (pid=fork ())
>>> +     {
>>> +     case -1:
>>> +       error (0, errno, "Could not run %s", path); @@ -63,7 +65,9 @@
>>> + 	  close (p[1]);
>>> + 	}
>>> +       dup2 (1, 2);
>>> +-      execve (path, argv, envp);
>>> ++      while (*envp)
>>> ++        putenv(*envp++);
>>> ++      execv (path, argv);
>>> +       _exit (127);
>>
>> What's that about?
>>
>> p.
> 
> Hi,
> 
> I believe the vfork() changes are for qemu compatibility - I don't know what the issue was exactly.
> 
> I didn't realise the linker version was for EABI/OABI compatibility - I guess prelink should determine the ABI and choose accordingly... 
> But I don't have any OABI binaries on my system at all...
> So we can:
> 	1. Fix it properly
> 	2. Switch from OABI only to EABI only
> 	3. Make a prelink-eabi recipe.
> 
> I don't really have the patience for 1., since I have no OABI binaries. Does anyone have any preferences here?
> 
> Regards,
> James
> 
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel