1 |
Being the user in question, I thought I'd clarify a few things |
2 |
|
3 |
|
4 |
On Fri, 19 Nov 2004 00:10:15 +0100, Carsten Lohrke <carlo@g.o> wrote: |
5 |
> On Thursday 18 November 2004 23:42, kosmikus wrote: |
6 |
> > We have had reports from users who tried to build this package on their |
7 |
> > 256Mb machines that the ebuild fails (no surprise there) and that the |
8 |
> > OOM killer takes out X in the process. |
9 |
> |
10 |
> Got a similar bug report. But hey, people know that you have to compile from |
11 |
> source, when using Gentoo. I don't think taking care for every <=500MHz box |
12 |
> low on ram does make sense. Usually you compile on another box, if you have |
13 |
> such a target. |
14 |
> |
15 |
> |
16 |
> Carsten |
17 |
> |
18 |
> |
19 |
> |
20 |
|
21 |
Since when is a machine with 256MB "low on ram"? (it's also a P4, |
22 |
though only 1.8GHz, but that's not really relevant) I can usually |
23 |
compile things whilst listening to music and not lose any |
24 |
repsonsiveness most of the time. Running out of memory is also |
25 |
different from having a slow processory - most people probably |
26 |
wouldn't realise that the reason that X had been killed was that |
27 |
they'd run out of RAM (especially considering the problems caused by |
28 |
gcc optimizations not so long ago). |
29 |
|
30 |
|
31 |
On Thu, 18 Nov 2004 22:55:42 +0000, Ciaran McCreesh <ciaranm@g.o> wrote: |
32 |
|
33 |
> Naah, I'd go for the warning with maybe an ebeep. If you really must |
34 |
> have a force thingie, it shouldn't be a USE flag, since USE flags are |
35 |
> for things which affect the resultant program |
36 |
|
37 |
The idea isn't to have a USE flag, but to do something like in the |
38 |
xorg-x11-6.8.0-r{2,4} ebuilds: |
39 |
|
40 |
pkg_setup() { |
41 |
################################################# |
42 |
### GET RID OF THIS ONCE THIS EBUILD IS READY ### |
43 |
################################################# |
44 |
if [ -z "${BREAKME}" ]; then |
45 |
die "Set the BREAKME variable to emerge this. It's in development. |
46 |
Stop using it." |
47 |
fi |
48 |
|
49 |
Except that it only fails if they don't have enough memory and if the |
50 |
variable isn't set. It's a nasty solution, but it can potentially stop |
51 |
bug reports about X randomly being killed. (After all, there's a high |
52 |
chance that someone won't notice any other sort of warning, beep or |
53 |
no) |
54 |
|
55 |
> I'd also be veeeeeeery wary about looking for a certain amount of RAM. |
56 |
> Chances are that the value in question will vary depending upon machine |
57 |
> and arch. |
58 |
|
59 |
Agreed, but if you give the user the option to still try the emerge |
60 |
then it's not so bad. Best solution out of a bad lot IMO... |
61 |
|
62 |
-- |
63 |
gentoo-dev@g.o mailing list |