This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RPATH/RUNPATH issue, equivalent to -headerpad on OSX


On Thursday 03 January 2008, Alexander Neundorf wrote:
> I am working on the build system for KDE4 and in KDE we are relying a lot
> on RPATH/RUNPATH.

kind of sad how every solution you talk about here is already taken care of 
transparently by libtool ... but at any rate ...

> There is a tool called chrpath available which does that:
> http://linux.die.net/man/1/chrpath
> http://websvn.kde.org/trunk/kdesupport/chrpath/
>
> But this tool has a problem, it only works if the new RPATH is not longer
> than the RPATH which is already in the ELF file (since it basically just
> patches the binary).

actually, that isnt its only problem.  imnso, chrpath is a quick hack that 
works for $author but does not scale at all beyond that.  it only works on 
ELFs that match the system you're building on, so you can forget about 
cross-compiling or even multilib (when bitsize is different).  either you 
move chrpath in house to make it not limited, or you screw your users by not 
allowing them to cross-compile or build multilib anymore.

that is why in Gentoo, we abandoned chrpath and instead used `scanelf -X`.  
scanelf handles the target endianness/bitness properly.  scanelf however does 
not dynamically adjust the string size accordingly either, it just truncates 
RPATH/RUNPATH down to the minimum.  perhaps it could be extended, but there 
was no such need/demand in Gentoo for it.

> Since a few days CMake (http://www.cmake.org) supports chrpath, i.e.
> instead of linking again when installing chrpath is used to modify the
> RPATH inside the binaries. A workaround for the mentioned problem is
> implemented here: if the RPATH for the install location is longer than the
> RPATH for the build location, then the RPATH for the build location is
> padded with separator characters. This works.
> But of course it would be much nicer if there was an official way to do
> this.

i hope cmake is fully aware of chrpath's limitations and documents it properly 
before people start relying on it.

> On OSX there is the -headerpad <size> option for ld, which basically does
> just that. So if you know that the RPATH for the install location is 200
> byte long, then you use -headerpad 200 and the linker will make sure that
> the field for the instll name will be at least 200 byte long.
>
> It would be nice if the ELF ld would offer a similar option,
> e.g. -rpath-length=<size>, and then the RPATH field would have the
> specified size. This would make it possible to safely modify the RPATH
> without workarounds.
> Maybe this would even make sense for binary distribution packages, where
> (maybe) the install location could be selected and then the RPATH inside
> the binaries of the package could be modified.

seems like this is a stopgap measure and not a real long term solution ... but 
who knows, maybe such a stopgap hack will work for your issue.
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]