more on ARM ELF flag merging
Nick Clifton
nickc@redhat.com
Wed Oct 25 13:29:00 GMT 2000
Hi Phil,
: I just discovered that ld was rejecting an attempt to interlink
: -msoft-float and -mhard-float objects without explaining why.
Oops!
: The patch below fixes this. and also rearranges things slightly so
: that the linker will produce messages for all the mismatches that
: exist, rather than just the first one.
A good idea.
One thing though - why does your patch remove the quick test to see if
the input flags and output flags are the same, and so return true
straight away ?
: Linking PIC code into a non-PIC executable is something that happens
: all the time on Linux (libgcc.a is PIC, for example) so we shouldn't
: generate an error or even a warning for that case. I think my
: preference would be to just silently clear the PIC flag on the
: output BFD if any non-PIC BFDs are merged with it; at the moment it
: inherits from the first input BFD that has any non-default flags.
: Nick, what do you think?
Well I would be OK with it. Ideally I would like to see a way to
enable these suppressed messages by a command line option. Originally
I thought of enabling them with --verbose, but this flag does not
appear to be passed through into the BFD library.
: 2000-10-22 Philip Blundell <philb@gnu.org>
:
: * elf32-arm.h (elf32_arm_merge_private_bfd_data): Print an error
: message for EF_SOFT_FLOAT mismatches. Display diagnostics for all
: mismatches, not just the first one.
Approved.
Cheers
Nick
More information about the Binutils
mailing list