| 1 |
> On 25 Jun 2022, at 09:34, Agostino Sarubbo <ago@g.o> wrote: |
| 2 |
> |
| 3 |
> Hello all. |
| 4 |
> |
| 5 |
> This is to make you aware that, per sam request, tinderbox is testing the tree against lto. |
| 6 |
> |
| 7 |
> At the time of writing, the CFLAGS/CXXFLAGS tested are: |
| 8 |
> -flto -Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing |
| 9 |
> |
| 10 |
> They are mentioned in the comment 0 of each bug. |
| 11 |
> |
| 12 |
> To make it more explicit, '(lto)' will appear in the summary too. |
| 13 |
> |
| 14 |
> The buglist is available here: https://tinyurl.com/yc4tu3cj |
| 15 |
> |
| 16 |
|
| 17 |
Thanks ago. |
| 18 |
|
| 19 |
From my perspective, the important thing here is to identify packages with serious quality |
| 20 |
issues and then filter-lto (which should be added to flag-o-matic.eclass shortly) and/or |
| 21 |
force -fno-strict-aliasing where appropriate. |
| 22 |
|
| 23 |
Maintainers aren't expected to dive in and fix these themselves. Depending on the |
| 24 |
status upstream, please filter-lto (and possibly append relevant flags depending |
| 25 |
on the error), and if it's alive, report upstream. |
| 26 |
|
| 27 |
Of course, feel free to fix it if you feel like. But the first step here is to avoid |
| 28 |
silent bad runtime behaviour. Hence filtering is not a sin. |
| 29 |
|
| 30 |
(Also, none of of the -Werror=* failures are going to be issues which |
| 31 |
only show up with LTO anyway.) |
| 32 |
|
| 33 |
Best, |
| 34 |
sam |
Archives