This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC 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: [PATCH] Fix -print-multi-os-directory for aarch64


On Fri, Mar 15, 2013 at 02:07:12AM +0800, Matthias Klose wrote:
> sorry, didn't comment about this patch because it didn't seem to affect
> multiarch.  However this patch assumes that every system does have at least a
> */lib64 symlink, if it doesn't have a */lib64 directory.  I think that is a
> wrong assumption.  Things like

Why do you think the patch assumes it?  And why do you care for multiarch?
It just adjusts where the libraries are searched.

In a cross, I see (for non-multiarch):
grep '^  \$ \.\.' ../config.log; strace -v -s1024 ./xgcc -B ./ -print-file-name=libc.so 2>&1 | grep access.*libc.so
  $ ../configure --target aarch64-linux --enable-languages=c,c++
access("./aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("./libc.so", R_OK)               = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa/gcc/../lib/gcc/aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa/gcc/../lib/gcc/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/../lib64/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/libc.so", R_OK) = -1 ENOENT (No such file or directory)

for multiarch:
grep '^  \$ \.\.' ../config.log; strace -v -s1024 ./xgcc -B ./ -print-file-name=libc.so 2>&1 | grep access.*libc.so
  $ ../configure --target aarch64-linux --enable-languages=c,c++ --enable-multiarch
access("./aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("./aarch64-linux-gnu/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("./libc.so", R_OK)               = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux-gnu/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/aarch64-linux/4.8.0/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/aarch64-linux-gnu/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/../lib64/libc.so", R_OK) = -1 ENOENT (No such file or directory)
access("/usr/src/gcc/objaa2/gcc/../lib/gcc/aarch64-linux/4.8.0/../../../../aarch64-linux/lib/libc.so", R_OK) = -1 ENOENT (No such file or directory)

as you can see, if lib64 paths don't exist, it just goes on to look in */lib (like it does on x86_64 etc).
And, for multiarch, it looks first in multiarch directories.

Checking for lib64 directory or symlink is what I'd prefer to avoid.

	Jakub


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