I ended up blowing a DTS incident and George Warner provided the help
that led to a solution.
For the benefit of Googlers Of The Future, I'm providing a synopsis of
what DTS suggested, and a description of what turned out to solve my
problem.
What the error message means: the PowerPC unconditional branch is 24
bits and the conditional branches are 14 bits.
How to diagnose more fully: extract the .o from the .a file using the
ar command, then run otool -rv on the .o file. Also run nm -n on the .o
file to match up the symbols.
How to avoid the problem: using -mlong-branch can help.
In my case I had a conditional branch that was trying to branch over
more than 14 bits, and this branch was in raw assembly code. (If it was
in C code, then the -mlong-branch flag would probably have helped, but
since it was in raw assembly the bne wasn't translated to avoid the
14-bit branch.)
Specifically, I had an inline assembly function that included code like
this:
1:
...
bne 1
And the fix (!) was to change it to look like this:
1:
...
bne 1b
Adding the "b" is a compiler hint to branch backwards. Without the "b",
it seems that it was trying to branch to a different invocation of this
(inline) function, which was a long long ways away. With the "b" it
branches backwards to the correct label.
Anyway, dunno if this is of interest to anyone, but if there's even a
chance that this will help ease the next guy's frustration it'll be
worth it.
--Bob
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-dev/email@hidden
