Re: [HACKERS] Some cleanups/enhancements - Mailing list pgsql-hackers

Thomas G. Lockhart wrote:
>
> > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So
> > far no problems, however I noted some cleanups / enhancements which I
> > would like to do. Before I send you a bunch of patches I thought I'll
> > tell you what I'm planning to do. Please comment on my list and indicate
> > whether I should go ahead.
> >
> > - Fix all Makefiles so 'make dep' and 'make depend' work
> > - Fix all Makefiles so 'make clean' throws away the depend file
> > - Some other Makefile cleanups
>
> These all sound good. If there is a possibility of large breakage, wait
> until after v6.3.

> Some or all of these changes might not be appropriate for v6.3, since we
> are in beta testing and since they do not affect the current functionality.
> For those cases, how about submitting patches based on the final v6.3
> release?

After the messages I've read so far, I'll wait until after the final
release of 6.3 and try to do the patches one at a time, so there'll be
plenty of time :-) to review them.

> [snip]

> > - Add a template for linux-elf-586 with (optimized) code for a Pentium
> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).
> > Why not use template names similar to the output of config.guess (maybe
> > with some symbolic links)?
>
> Does gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,
> then the template should be more explicit in name (e.g.
> "linux-elf-586-gcc2.8") or we should update the FAQ or include comments in
> linux-elf with some suggestions. It was only in the last release or two
> that the -m486 was added, and I worried about causing trouble for _all_ of
> those 386 users :)

No, it doesn't. linux-elf-586-gcc2.8 sounds OK to me.

> [snip]


The Hermit Hacker wrote:
> > - Fix all Makefiles so 'make dep' and 'make depend' work
>
>         Can someone explain what, exactly, 'make depend' accomplishes?  We
> don't use it right now, so I'm wondering why (if?) we need it now?

It makes sure that your C files get compiled if you change a header
file. Your C compiler should be able to find out which files are
included and create lines which can be included in the Makefile (man gcc
:-) )

> > - Some other Makefile cleanups
> > - gcc 2.8.0 issues some additional warnings which are very easy to fix:
> >   - register i --> register int i
> >   - Ambiguous else --> add braces:
> >     if (cond1)
> >       if (cond2)
> >         ...
> >       else
> >         ...
> >   - etc.
>
>         Sounds great...

If I find something like this, I'll remove the register as well.

> > - Add a template for linux-elf-586 with (optimized) code for a Pentium
> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).
> > Why not use template names similar to the output of config.guess (maybe
> > with some symbolic links)?
>
>         Erk...I think 'templates' are getting a little out of hand here,
> no?
>
> > - Why is there some code to change the case of the procedural language
> > to lower case except for 'C' (in fact it's there twice)? Why not use
> > strcasecmp and remove these pices of code?
>
>         I question this in *alot* of places...like why pg_dlopen is
> defined as just 'dlopen()' in some ports *shrug*  Why not just call it
> directly? *raised eyebrow*
>

[snip]

Bruce Momjian wrote:
>
> I recommend running the regression test before/after the changes, to
> make sure something didn't get broken.

Sure.

> [snip]

See you in a 2-2.5 weeks :-)

Jeroen van Vianen