1 |
On Sat, 2006-08-05 at 12:57 +0200, Kevin F. Quinn wrote: |
2 |
> On Sat, 5 Aug 2006 11:49:53 +0200 |
3 |
> Danny van Dyk <kugelfang@g.o> wrote: |
4 |
> |
5 |
> > Please re-read the list of packages that fail tests: |
6 |
> > * glibc |
7 |
> > * autoconf |
8 |
> > * gettext |
9 |
> > * tar |
10 |
> > That makes _4_ system packages. Before I would consider making |
11 |
> > FEATURES=test a default, I would add least want the system set to |
12 |
> > actually merge with it. |
13 |
> |
14 |
> So you're happy to let users install these packages without them |
15 |
> knowing the tests would fail? |
16 |
> |
17 |
> I certainly agree they should pass their tests. autoconf-2.60, |
18 |
> gettext-0.15 and tar-1.15.1-r1, which are the latest versions I |
19 |
> have installed here, all pass on my system. If they fail on your |
20 |
> platform, then you should make sure bugs are open and the relevant |
21 |
> maintainers are doing something about it, and IMO they should not go to |
22 |
> arch (i.e. should remain ~arch) until the test issues are resolved. |
23 |
> |
24 |
> Thing is, at the moment you have a bunch of packages installed that |
25 |
> fail their tests. This may mean the tests are broken, however it may |
26 |
> also mean the packages are not working correctly on your system, and |
27 |
> I'd be concerned if I were you. |
28 |
|
29 |
With some arches this is not really an option. Also system pkgs such |
30 |
like the toolchain need to have additional deps. |
31 |
|
32 |
> Avoiding the test phase doesn't make |
33 |
> the packages work, obviously. |
34 |
> |
35 |
> glibc is somewhat of a special case; it is especially sensitive to |
36 |
> the environment - many of the tests assume a vanilla RedHat |
37 |
> environment, and often the test failures in glibc are not actual |
38 |
> problems with glibc but limitations of the test suite. |
39 |
|
40 |
Sometimes the tests are flat out wrong. |
41 |
Take for example say we decided to paxtest ran itself in as the test.. |
42 |
This would surely fail on amd64 as one or two of the tests assume page |
43 |
sizes of 4096. |
44 |
|
45 |
> However we |
46 |
> should not be encouraging people to install glibc versions where the |
47 |
> test failures are not understood. |
48 |
|
49 |
The alternative would then become for the end user to use |
50 |
another distro with less hassles. We would surely get the rep |
51 |
of sucking if nobody could even install libc. |
52 |
|
53 |
> Clearly if something in glibc is not |
54 |
> behaving properly, the effects can be nasty. |
55 |
|
56 |
Which for the most part is why features like |
57 |
this should be opt-in vs opt-out or be left up |
58 |
to the $ARCH teams. |
59 |
|
60 |
A lot of people are opting in so most of these will be |
61 |
fixed in due time.. The $ARCH teams *should* already be setting |
62 |
this feature for the most part before stable markings. |
63 |
|
64 |
It's a noble idea. I just don't think we are ready for |
65 |
FEATURES=test && USE=test either. |
66 |
|
67 |
|
68 |
-- |
69 |
Ned Ludd <solar@g.o> |
70 |
Gentoo Linux |
71 |
|
72 |
-- |
73 |
gentoo-dev@g.o mailing list |