| 1 |
On 12/03/14 19:11, Michał Górny wrote: |
| 2 |
> Hi, |
| 3 |
> |
| 4 |
> Some observations I made when forking gcc. Hope you find them helpful |
| 5 |
> for toolchain.eclass work. |
| 6 |
> |
| 7 |
> 1. src_unpack() is likely unnecessary -- the default will unpack |
| 8 |
> the fetched tarballs, |
| 9 |
> |
| 10 |
> 2. setup_multilib_osdirnames() doesn't do anything with gcc-4.9 |
| 11 |
> (possibly some older too) -- the sed expression doesn't match anything, |
| 12 |
> |
| 13 |
> 3. *-linux) needed_libc=no-fucking-clue;; |
| 14 |
> very professional, isn't it? |
| 15 |
> |
| 16 |
> 4. if you enable --enable-version-specific-runtime-libs and add a small |
| 17 |
> patch from funtoo, you can get rid of that big post-install library |
| 18 |
> moving thing. |
| 19 |
> |
| 20 |
> 5. the 'gdbdir' definition fails on non-SYMLINK_LIB systems -- i.e. |
| 21 |
> when gcc is actually in /usr/lib/gcc. |
| 22 |
> |
| 23 |
> 6. some things just ask for REQUIRED_USE rather than silent disabling. |
| 24 |
> |
| 25 |
> 7. you miss dep on doxygen with USE=cxx,doc. |
| 26 |
> |
| 27 |
Thanks for posting this. Can I please see where you are working on this |
| 28 |
so I can test and make sure this is going to work with the stuff I do |
| 29 |
--- minor arches + alternative/embedded libc's. I'd like to pre-empt |
| 30 |
issues and not repeat some of the headaches I hit with multilib because |
| 31 |
I got in on the act late. Also, I have a series of commits I'd like to |
| 32 |
get into toolchain and have held off ---- again to do with alt libc's. |
| 33 |
|
| 34 |
-- |
| 35 |
Anthony G. Basile, Ph.D. |
| 36 |
Gentoo Linux Developer [Hardened] |
| 37 |
E-Mail : blueness@g.o |
| 38 |
GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA |
| 39 |
GnuPG ID : F52D4BBA |
Archives