Re: [5.2.1] quiet warnings for gdbreplay.c

[Andrew Cagney <ac131313@ges.redhat.com>]

> I dunno, I appreciate being able to run configure in the gdbserver/
> directory.  It means you don't need to a terminal library for your
> target if you just want gdbserver.  A lot of people have seemed to
> appreciate this.

Ah, ok, I see your problem.  Yes that does make things easier. 
GDBserver on its own is far more likely to configure/build and is easier 
to drag around.

Suggest a comment (I've no idea where) explaining this (or even mention 
it in GDB's doco - something like stating that gdbserver can be 
built/configured separate to GDB).

>> More seriously, I think someone hacking on gdbreplay is likely to also 
>> be hacking on GDB.  Consequently, they are going to expect the two 
>> directories to follow the same coding conventions.
> 
> 
> We discussed this.  The two directories share no code except for
> signals/signals.c off in its own space.  If they share headers, some
> care is called for.  They are not part of the same program!

True, but they are part of the same ``system''.

>>   I had
> 
>> >eliminated all references to the gdb source code but then I introduced
>> >an include of "gdb_proc_service.h", since the alternative was just
>> >duplicating it; I have the feeling we should move that and headers like
>> >gdb_string.h somewhere common - are they include/gdb/ candidates?
> 
>> 
>> You mean #include "gdb/gdb_string.h"?  I think include/gdb/ is for 
>> external interfaces that are at some level controlled by GDB.
> 
> 
> Then do we want a separate gdb/gdbint/ directory for this?  I strongly
> prefer that headers shared between GDB and other directories be clearly
> marked and separated.  That'd give me a place to move
> gdb_proc_service.h, too.

I think it would be easier to just clarify the guidelines for 
``gdb_XXXX.h'' files - that they be independant as they are included by 
GDB and friends.

In the mean time, I guess the status quo remains.

Andrew