This is the mail archive of the gdb@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: New ARI web page, generated using script inside CVS tree in gdb/contrib/ari directory


> 3) stat.h	1	Do not include stat.h or sys/stat.h, instead include
> gdb_stat.h
> 4) wait.h	1	Do not include wait.h or sys/wait.h, instead include
> gdb_wait.h
[...]
> 3) and 4) are all related to the fact that gdb_stat.h and gdb_wait.h
> headers are still in gdb directory, which force the use of direct
> system include headers in sources in gdb/common sub-directory.  Those
> can probably be solved easily by moving the two headers also to
> gdb/common.

I had a look at gdb_stat.h and gdb_wait.h, and I think it would be
fine to move them over to gdb/common.

But perhaps we should also look at possibly importing gnulib's stat
and sys_wait modules? Not a trivial change in the sense that it could
be not equivalent to what we have now, and thus have unintended
consequences; but perhaps worth a shot. I think that gdb_stat.h and
all other such headers were GDB's own way of doing what gnulib does
in general.

> 1) is a little bit more tricky because safe_strerror is declared in utils.h
> header,
> but implemented in two files:
> posix-hdep.c and mingw-hdep.c
> 
>   Should I extract the declaration into a
> gdb/common/gdb_strerror.h
> and extract the two functions
> into gdb/common/posix-strerror.c and gdb/common/mingw-strerror.c?

This sounds good to me. There is the option of putting all
implementation into one single file, the way we do for some
of our code, but I don't think it would be particularly simpler
to do that way, since the current code has it separated in 2 files.

In terms of the name, perhaps we should make it more explicit that
this is not strerror, but safe_strerror?

> Should we add rules about
>   gdb_XXX.h headers
> stating that if gdb_XXX.h header exists and shadows an existing
> system header, no gdb C source file should include XXX.h system header
> directly.

I would think so - but we should also look at gnulib, to see if
it couldn't just be replaced by gnulib's equivalent.

>   Also, I would find as a logical consequence that also
> gdbserver subdirectory should follow the ARI rules.
>   This is done quite easily by removing the 
>   -name gdbserver -prune -o 
> line from gdb_find.sh script in gdb/contrib./ari
> but is a rather important change that should be discussed fully.

Right. I think it would be a helpful, but I'll let Pedro comment
on that one.

-- 
Joel


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]