Re: build improvement

[Ben Elliston]

>   Tangled tree of Makefiles ? Other than the examples, there are
> very few Makefiles. I've played with the single top level Makefile
> idea before, and it is faster. Whether it's more maintainable or not
> I think is personal opinion. I don't have any motivation to make
> this change myself, but if somebody else did, I probably wouldn't
> gripe when it came time to check it in. Most of the DejaGnu
> Makefiles are pretty simple.

Here's the list:

./example/calc/Makefile.am
./example/hello/testsuite/Makefile.am
./example/hello/Makefile.am
./example/Makefile.am
./example/mathhelper/libmathhelper/Makefile.am
./example/mathhelper/Makefile.am
./example/mathhelper/testsuite/mathhelper/Makefile.am
./example/mathhelper/testsuite/Makefile.am
./doc/C/Makefile.am
./doc/Makefile.am
./testsuite/libdejagnu/Makefile.am
./testsuite/Makefile.am
./Makefile.am

You're right -- some of them are simple: they just have a SUBDIRS
variable, which makes this even easier to collapse.

>   Just as a question, what other projects ? I haven't downloaded a
> single free software package using this technique, so I'm curious.

libjava.  It's wonderful to watch `make' sit there for 10 seconds
working out what it really needs to do rather than forking off
sub-makes that do nothing.

Ben

pgptifzyJufC7.pgp
Description: PGP signature