Re: What has happened to IN_SOLIB_CALL_TRAMPOLINE?

[Andrew Cagney <cagney at gnu dot org>]

Mark Kettenis wrote:
> Any calls to this macro or the equivalent
> gdbarch_in_solib_call_trampoline seem to have disappeared.  What's
> happened to it?

Dig dig, best read is the #if 0 code in gdb 4.18.

It was part of a long complicated sequence of tests used to detect 
step-into-function.  Working unwind code (in particular, a willingness 
to trust frame IDs) made it all redundant.

Architectures with stub-unwinders had an equivalent test (in plt), and 
in the case of hppa, the only remaining call.

> Many targets still set it, but of we can really do without it, we
> shdould probably go over the sources and remove a lot of old junk.

gone

> I'm not sure that we can entirely do without some sort of shared
> library call stub skipping mechanism.  On most targets these stubs are
> frameless and won't show up in backtraces.  But I'm not completely
> sure if that's true for all targets.  Currently I'm looking at mips,
> and hppa should be investigated too.
>
> I'm fairly sure that amd64, i386, sparc and sparc64 can do without.

Andrew