Wrong compiler for discord_codecs breaks cross about concord

[HackerSmacker]

A guide has been added, and, it seems pretty streamlined for new users. Nonetheless, we don't target GNU Make, and some Make implementations that Concord builds under (e.g. SVR3 Make) can't be expected to have some variables predefined. Conditional assignment is flakey on some Makes, and, systems with non-GCC compilers won't take the conditionally-assigned CFLAGS anyways -- we try to bend over backwards to make that Makefile work on any make implementation out there on a POSIX.2 system with a C99 compiler, so, there's a fair bit of weirdness in there.

If you run into any more issues, feel free to give us a shout -- good luck on your project!

[HackerSmacker]

The current procedure is to enter the gencodecs directory and invoke make with $(CC) overriden on the command line or in the Makefile, however, this does NOT properly cross-compile. Let me fix that and make a few notes on that!

[HackerSmacker]

Okay, it should be fixed. I made a note in the README that explains how to do it -- I tried to make it as simple as possible. Feel free to drop another comment if you've got any more feedback, and thanks for the report!

[OPNA2608]

Hmm, well cross-compiling works perfectly well for me if I patch that Makefile to use a non-CC variable and inherit CC from the main Makefile that calls into it instead. I feel like having a separate step for gencodecs is just pointless complexity.

[HackerSmacker]

Well, to be fair, there is a reason the architecture is the way it is -- gencodecs is not a component that is exclusive to Concord, and was designed to be somewhat generic. Regardless, the change has been made and it should be pretty easy for 99.99% of users to roll with.

[OPNA2608]

gencodecs […] was designed to be somewhat generic

I think that's even more of an argument in favour of not needlessly overriding these implicit Make variables then, or changing something about the way these are assigned at least. With GNU Make, CC and CPP are already implicitly defined and have reasonably-likely fallback values. The file already uses conditional variable assignment for CFLAGS, so the same could be done for those two if you really want to be sure that it has some assignment.

But yes, feel free to close as the original issue has been resolved. Just trying to cut down the patching I need to do so everything behaves as it should.