toolchain.eclass: stop building JIT separately

We introduced a split for libgccjit because it needs --enable-host-shared
for bug #843341, but I think this was a mistake in the end:
* We don't bootstrap that first, host-shared build because otherwise the
  build would take even longer;

* gcc PR117047 is an example where not-bootstrapping causes very-hard-to-debug
  problems because libgccjit may be compiled differently by a newer compiler,
  but you can't just rebuild GCC once to observe that. Even knowing this*,
  I was stumped by it for some time;

* It introduces complexities into the ebuild and it's already complex enough.
  See bug #954077 and bug #953823;

* We want to support rust_codegen_gcc (a rustc codegen backend using libgccjit)
  in the future. If we do that, we want a bootstrapped build for libgccjit
  to know it's reliable;

* The test setup as-is doesn't run tests for the first build;

* On slower machines, that first "wasted" build is pretty noticeable
  and slow.

I think any possible slowdown should be mitigated by using LTO and possibly
-fno-semantic-interposition for GCC, which users are free to do.

Bug: https://bugs.gentoo.org/843341
Bug: https://bugs.gentoo.org/954077
Bug: https://bugs.gentoo.org/953823
Signed-off-by: Sam James <sam@gentoo.org>

0 files changed, 0 insertions, 0 deletions