configure: handle paths with spaces in getExecutable(). Now the following works:
'--with-mpiexec=/cygdrive/c/Program\ Files/MPICH2/bin/mpiexec -localonly',

make: fix usage from programs.make to self.make [via packages.py]

make: clean up integration of programs.make into make.py

download-make: migrate prior 'make' detection code from programs.py to packages/make.py

download-make: add initial support. [yet to migrate stuff from programs.py]

config: dependency flags are only needed by compiler, so skip linker test

config: add dependency flags for most compilers

download MOAB with git if possible
fix missing ierr when setting moab citations

config: fix classification to return ALL good/bad items

gmake: use configure test for -MMD in gmakefile

config libraries: skip link test when functions=[]

This is needed for ctetgen, which depends on petscconf.h, thus cannot
run before configure completes.

Revert "config BlasLapack: remove unused checkForRoutine"

This function is indeed used, just not by BuildSystem.

This reverts commit d24de7f2040708be0b5e7710639a27820a292bd6.

config: add @memoize decorator and use for setCompilers.isGNU

config: remove unnecessary unused variable in test

config: consolidate classification based on boolean set functional

Eliminates code duplication caused by having essentially same "advised"
bisection logic in functions.py and libraries.py.

The comm

config: consolidate classification based on boolean set functional

Eliminates code duplication caused by having essentially same "advised"
bisection logic in functions.py and libraries.py.

The common functionality is stateless, thus resides in its own module.

show more ...

config MPI: use libraries.checkClassify to find missing functions

config functions: change ENOSYS safeguard to proceed further

Use of #error causes compilation to bail early without identifying all
the problem functions. This will still cause an error, but will m

config functions: change ENOSYS safeguard to proceed further

Use of #error causes compilation to bail early without identifying all
the problem functions. This will still cause an error, but will make it
further likely identifying all problem functions in the error message,
thus allowing faster batch classification.

show more ...

config: add examineOutput callbacks to optimize functions.checkBatch

Missing symbols are usually named in stderr when a compilation or link
fails. That information can be used to select those funct

config: add examineOutput callbacks to optimize functions.checkBatch

Missing symbols are usually named in stderr when a compilation or link
fails. That information can be used to select those functions that are
more likely responsible for the compilation failure.

On my machine, this speeds up config.functions from 11 seconds to 5 seconds.

show more ...

config functions: add checkClassify

Optimistically check for many functions. Currently not faster because
the default function lists have too many likely failures.

config BlasLapack: mangleBlas does not need special logic for f2c

config BlasLapack: trivial loop clean-up

config BlasLapack: remove unused checkForRoutine

config: add libraries.checkClassify to find missing functions, use in BlasLapack

Usually we find most of the functions that we need to look for, so be
optimistic and try to compile and link them all

config: add libraries.checkClassify to find missing functions, use in BlasLapack

Usually we find most of the functions that we need to look for, so be
optimistic and try to compile and link them all at once. If that fails,
recursively bisect so that we usually perform less than one compile/link
per function that we need to check for.

Using this in BlasLapack cuts BlasLapack search time in half.

show more ...

config libraries.check: test multiple functions in single compile

config libraries.check: consolidate function-specific parts

This is in preparation for recursive batch checking for functions.