r147958 - in /trunk/libstdc++-v3: ChangeLog Mak...

[Tom Tromey]

>>>>> "Joseph" == Joseph S Myers <joseph@codesourcery.com> writes:

Tom> * It now computes the new module path relative to the objfile.  This
Tom> is better, in that moving the whole tree should work.  However, it
Tom> assumes that datadir is always available as $libdir/../share.
Tom> I'm not sure whether that is ok.

Joseph> There are two problems:

Joseph> * You can't assume it's $libdir/../share.

Ok.

Joseph> * Shared libraries are naturally copied from one place to
Joseph> another, with the -gdb.py file copied along with them.  In
Joseph> particular, for a cross compiler to a GNU/Linux target they
Joseph> would be copied to the /usr/lib directory in a sysroot
Joseph> directory that is the image of the target's root filesystem
Joseph> and may be referenced with --sysroot in compilation commands,
Joseph> "set sysroot" in GDB, etc.; in this case, do you want to
Joseph> require all the share/ files (which may only be of use to a
Joseph> cross-debugger, not on the target) to be copied there as well?

I don't follow.  Are you saying you want to copy just the -gdb.py file
but nothing else, and have it work?  I don't think that is a scenario
we should support.  At least in theory, the python code is specific to
a given version of libstdc++.  So, if you want to use the pretty
printers, you should copy them around too.  Furthermore, I think we
ought to reject host- or target-specific changes to the python code.
If we need such code, we should find a way to make gdb expose the
needed information instead.

Tom