1 |
On 1/18/15 10:50 PM, Anthony G. Basile wrote: |
2 |
> Hi everyone, I'd like to make a commit to toolchain.eclass in a few |
3 |
> days. mgorny noticed some code which can be improved. Basically gcc |
4 |
> creates "fixed" include files from system headers because of the |
5 |
> requirement that it have ansi c compliant headers. These are fixed via |
6 |
> shells scripts during the build process, but we just delete them |
7 |
> afterwards. Rather than generate and then delete them, just don't |
8 |
> generate them in the first place. Its bug number #536878. I've tested |
9 |
> on gcc-3 to the latest and both approaches yield the same results. |
10 |
|
11 |
Sounds good. |
12 |
|
13 |
> Here's the patch so you don't have to go hunting for it: |
14 |
> |
15 |
> diff -u -B -r1.647 toolchain.eclass |
16 |
> --- toolchain.eclass 15 Nov 2014 08:45:33 -0000 1.647 |
17 |
> +++ toolchain.eclass 18 Jan 2015 20:36:08 -0000 |
18 |
> @@ -595,6 +595,13 @@ |
19 |
> einfo " ${f%%...}" |
20 |
> done |
21 |
> fi |
22 |
> + |
23 |
> + # we don't want fixed includes :) |
24 |
|
25 |
Please consider summarizing above in the comment (mostly to the effect |
26 |
that fixed headers are deleted afterwards and are not needed on modern |
27 |
systems). It's obvious it's just disabling the logic in given shell |
28 |
scripts, but from the code itself it's not obvious why. |
29 |
|
30 |
Paweł |
31 |
|
32 |
> + if tc_version_is_at_least 4.0; then |
33 |
> + echo : > "${S}"/fixincludes/fixinc.in || die |
34 |
> + else |
35 |
> + echo : > "${S}"/gcc/fixinc/fixincl.sh || die |
36 |
> + fi |
37 |
> } |
38 |
> |
39 |
> guess_patch_type_in_dir() { |