| 1 |
On Monday 21 September 2009 18:16:32 Paige Thompson wrote: |
| 2 |
> I hope nobody finds this offensive, I'm not a great writer but I gotta get |
| 3 |
> this out there. |
| 4 |
> |
| 5 |
> Goal: to resolve quality issues with packages and the behavior of portage |
| 6 |
> |
| 7 |
> Problem 1: |
| 8 |
> |
| 9 |
> This is a really simple thing, first of all it would help a lot if packages |
| 10 |
> will not try to build with specified cxxflags if the maintainer hasn't |
| 11 |
> tested the build and enabled them for that package. |
| 12 |
> |
| 13 |
> case and point: |
| 14 |
> I have -fstack-protector-all in my cxxflags because I'm a paranoid idiot |
| 15 |
> and I'm overly confident that it could never be wrong to have that. emacs, |
| 16 |
> fails to build because of it but it's not obvious. I file a really |
| 17 |
> pedantic bug report, and later through trial and error and after having |
| 18 |
> gotten over my confidence in -fstack-protector-all realized that without |
| 19 |
> it the package *does* build. If the ebuild had a feature where it's |
| 20 |
> metadata did not indicate that it could build with that cxxflag, then |
| 21 |
> portage could stop and just tell me that up front *OR* prompt me and ask |
| 22 |
> me what do next. I understand that this would require package maintainers |
| 23 |
> to actually *test* their packages which is no trivial issue, and who |
| 24 |
> wouldn't agree that if they're not willing to then somebody else should? |
| 25 |
> Not only that but it gives you the ability to score maintainers based on |
| 26 |
> the accuracy of the results. I'm not even suggesting that this feature |
| 27 |
> should be mandatory it could be something that I could turn on or off-- I |
| 28 |
> just want it so that I know what's going on and I don't end up wasting |
| 29 |
> people's time filing bug reports and making them mad at me for being a |
| 30 |
> noob. |
| 31 |
> |
| 32 |
> Problem 2: |
| 33 |
> I know this is might be kind of nitpicky to you, and it's more or less the |
| 34 |
> same as problem 1 but I think if I specify -O0 in my cxxflags, that a |
| 35 |
> package that needs -O2 should not build and tell me that it needs it rather |
| 36 |
> than just building with -O2 anyway!! I mean seriously, why even give me the |
| 37 |
> option to specify the optimization level in the cxxflags. It's deceptive, I |
| 38 |
> don't like that I find it very difficult to take it seriously because of |
| 39 |
> that. |
| 40 |
|
| 41 |
This is the wrong forum for that. You need to take it up with the devs, IIRC |
| 42 |
they are at gentoo-dev. Few of them read this list. |
| 43 |
|
| 44 |
|
| 45 |
-- |
| 46 |
alan dot mckinnon at gmail dot com |
Archives