Mailing List Archive

On Mon, 20 Nov 1995, Rob Hartill wrote:
> > What do people feel about this addition to the document about binary
> > builds:
> >
> > <B>Compiling:</B> The GNU C compiler (gcc) should be used whenever possible,
> > and with the -g flag set, so that we can ask people using binaries to provide
> > a backtrace if a coredump occurs.
>
> Hmmm, I get the impression that this statement would be aimed at people
> who don't really know their way around unix/C enough to run the debugger.
> It might just prove to be one more thing to confuse the hell out of
> a newbie.

Hmm, it's also probably reasonable to presume that someone without gcc is
also likely to be without a gdb too :) The performance penalty in that
case might not be worth it then.. can I get one or two more opinions on
the subject?

> Also, I think that if people have good compiler installations they
> shouldn't be persuaded to use gcc instead. By all means recommend gcc,
> but don't make it sound like it should be used in preference to the
> native CC. Since installing the HP compiler, I've stopped using gcc,
> and have fewer problems compiling applications.

If the debugging benefit is lost anyways, then people should be
encouraged to compile using whatever they consider the "best" compiler
for the platform, so I agree. Unless of course the compiler contains
licensing restrictions preventing its use on software to distribute for
free...

Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/

> What do people feel about this addition to the document about binary
> builds:
>
> <B>Compiling:</B> The GNU C compiler (gcc) should be used whenever possible,
> and with the -g flag set, so that we can ask people using binaries to provide
> a backtrace if a coredump occurs.

Hmmm, I get the impression that this statement would be aimed at people
who don't really know their way around unix/C enough to run the debugger.
It might just prove to be one more thing to confuse the hell out of
a newbie.

Also, I think that if people have good compiler installations they
shouldn't be persuaded to use gcc instead. By all means recommend gcc,
but don't make it sound like it should be used in preference to the
native CC. Since installing the HP compiler, I've stopped using gcc,
and have fewer problems compiling applications.

Just some opinions.

rob

Okay, here's the revised text:

<B>Compiling:</B> Use the "best" compiler for the platform, given that the
object code built with the compiler is free from any restrictions the
compiler might put on it. Set for the first or second level of
optimization, and don't compile with debugging since anyone who can use a
debugger to backtrace a core dump can probably also compile it themselves
anyways.

Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/

>
> On Mon, 20 Nov 1995, Rob Hartill wrote:
> > > What do people feel about this addition to the document about binary
> > > builds:
> > >
> > > <B>Compiling:</B> The GNU C compiler (gcc) should be used whenever possible,
> > > and with the -g flag set, so that we can ask people using binaries to provide
> > > a backtrace if a coredump occurs.
> >
> > Hmmm, I get the impression that this statement would be aimed at people
> > who don't really know their way around unix/C enough to run the debugger.
> > It might just prove to be one more thing to confuse the hell out of
> > a newbie.
>
> Hmm, it's also probably reasonable to presume that someone without gcc is
> also likely to be without a gdb too :) The performance penalty in that
> case might not be worth it then.. can I get one or two more opinions on
> the subject?

Also this brings up the subject of how much we trust our software. Do we
expect to have core dumps? I doubt we're going to get many coredumps, rather
annoying seutp questions, or misconfigurations.

Also, if who eer is looking for a binary disterbution. I doubt they would have any idea on how to do a backtrace.

<Aram>

Hmmm, I got confused. I though the info was for general help for
compiling Apache. I guess some of what I said is still relevant though.

> Okay, here's the revised text:
>
> <B>Compiling:</B> Use the "best" compiler for the platform, given that the
> object code built with the compiler is free from any restrictions the
> compiler might put on it. Set for the first or second level of
> optimization, and don't compile with debugging since anyone who can use a
> debugger to backtrace a core dump can probably also compile it themselves
> anyways.

Nope, this is for the people building the binary distributions.

Brian

On Mon, 20 Nov 1995, Rob Hartill wrote:
> Hmmm, I got confused. I though the info was for general help for
> compiling Apache. I guess some of what I said is still relevant though.
>
> > Okay, here's the revised text:
> >
> > <B>Compiling:</B> Use the "best" compiler for the platform, given that the
> > object code built with the compiler is free from any restrictions the
> > compiler might put on it. Set for the first or second level of
> > optimization, and don't compile with debugging since anyone who can use a
> > debugger to backtrace a core dump can probably also compile it themselves
> > anyways.
>
>
>
>

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/

> <B>Compiling:</B> The GNU C compiler (gcc) should be used whenever possible,
> and with the -g flag set, so that we can ask people using binaries to provide
> a backtrace if a coredump occurs.


Seems sensible.

> Brian

Ay.