| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
Attempt to pull in generic libraries whenever we see a platform
optimized one. This implementation uses the naive approach of just
chopping directories out of the middle of library paths and seeing if
there's a file there without that part. If there is, it pulls that in
as well.
It seems to work to build power5-able trees on a power6 machine.
|
| |
|
|
|
|
|
|
| |
We want a power6-generated initrd.img to be usable on power5 and such,
so in general we don't want get_dso_deps() to give us subarch-optimized
versions of libraries.
Resolves: rhbz#572178
|
| | |
|
| |
|
|
|
|
| |
building x86 boot images on a x86_64 host system.
Also don't try to verify the binary if $LDSO is empty.
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
This was causing get_dso_deps to go haywire when encountering
statically linked things like /sbin/grub. Particularly bad
on x86_64 when it would try to use the 64bit LDSO to run the
32bit executable. The output would be bad enough that it would
short circuit our loop to install all the things in KEEPFILES. We
noticed this because on x86_64 the gtk2 theme is what is listed after
/sbin/grub in keepfiles, and we weren't getting our theme on x86_64.
|
| |
|
|
|
|
| |
Exiting from a function being called could leave things in a
weird state. The exit makes sense for mkinitrd, maybe less
so when building installer images
|
| |
|
|
| |
If s390 images lack /lib/ld64.so.1, nothing works.
|
| |
|
|
|
| |
mk-images calls get_dso_deps as well, without a library argument. Since
everything there should be of the primary arch, just set a default.
|
| |
|
|
|
|
|
| |
- Call get_file_deps with a an argument that specifies the libdir based
on whether the binary is 32-bit or 64-bit.
- Check whether the currently cached $LDSO actually works for the binary
passed - if not, look for a new one in the passed libdir
|
| | |
|
| | |
|
| | |
|
| |
|
