compile time testsuite [Re: alias slowdown?]

[Benjamin Kosnik <bkoz at redhat dot com>]

> Perhaps we need a webpage with multiple compile times being tracked,
> or simply a standard compile time regression suite from existing PRs
> to avoid this kind of thing in the future.

I think it is past time for a formal compile-time testsuite.

My suggestion would be to check in 

all.i
bug2.ii
cpgram.ii

And whatever stuff there is for measuring slowdowns, and then create
an optional "make check-compile" rule for gcc/Makefile that runs
through the tests. I believe that you will have to make this testing
always optional, since there will no doubt be pathological behavior
involved.

Then you could do

time make check-compile

And get a better picture of what's up. I'm sure there are other ways
of doing this (better?), and I'd be interested in hearing them.

Regardless, I think it's important to start centralizing test cases in
the GCC source repository. It's crazy to think that gcc developers
have to data-mine gcc and gcc-patches to test for compile-time
issues...

FWIW, libstdc++-v3 has a "make check-compile" rule that just runs the
usual test suite, but with -g -O2 -S. It's been useful for
compile-time testing, for extra warning testing, and for non-GNU
compiler testing.

If something in this area gets standardized for gcc proper I'd like to
make sure that names/behavior stay somewhat similar.

best,
-benjamin