Re: patch to common.make for cygwin paths

[Stephen Brandon]

Hi,

This would be a lot easier if MAKE had full regex support
:-)

I think the warnings are good and I would not want to remove
them from Windows compilation, where, to be honest, it's
easier to make the mistakes because it can be harder to
ensure that GNUstep.sh has been run before doing
compilations etc. It's not as standardised as unix.

I understand about the performance implications, but bear in
mind that common.make generally only gets run once per
compile (at least, the bogus warnings were appearing for me
only once per compile).

If you want to surround my checks or the original checks
with checks for cygwin, please also check for mingw32
(that's the target that I work to, and the one I would
recommend for anyone, because of cygwin licensing
implications). [I think I used the word "check" too many
times there...]

It's good to have cygwin installed while compiling because
it provides lots of good tools, but working to a mingw32
target eliminates Cygwin.dll dependencies in the compiled
code.

Cheers,
Stephen

On Tue, 6 Nov 2001 10:58:54 +0000 (GMT)
 Nicola Pero <nicola@brainstorm.co.uk> wrote:
> hi,
> 
> I don't know if your patch actually does something useful
> because I don't
> know much about windows ... :-) I'll wait for Adam - who
> knows about
> windows - to rule on this matter_
> 
> but I want to address immediately the performance
> problems it might cause
> [on unix machines as well] ... running subshells and
> firing sed commands
> must be avoided in the processing mainstream ... I'd
> rather remove the
> check altogether if that is the only way of performing
> it_
> 
> we might either completely remove the check at all on all
> machines, or
> disable it on windows [and any other system where it is
> too
> computationally expensive], or you need to figure out a
> way to perform the
> check without running additional subshell commands_
> 
> the check is pretty irrelevant, and I added it because it
> was so
> computationally cheap on unix ... and hoped it might
> prevent some unaware
> people from complaining that the test service in
> gnustep-gui didn't work
> because they hadn't run the GNUstep.sh script ... or had
> run it but then
> messed so much with users that they no longer had the
> GNUstep tools in
> their path ... but I have nothing particular against
> completely removing
> the check if it is too expensive to perform_
> 
> my preferred solution would be to enclose the check into
> a 
> 
> ifneq ($(findstring cygwin, $(GNUSTEP_TARGET_OS)),
> cygwin)
> 
> endif
> 
> so that it is not performed at all on windows, and is
> performed only on
> systems were it is irrelevantly quick_ 
> 
> comments appreciated_
>