1 |
On Thu, 12 Feb 2009 09:10:55 -0800 |
2 |
Donnie Berkholz <dberkholz@g.o> wrote: |
3 |
> On 16:21 Thu 12 Feb , Ciaran McCreesh wrote: |
4 |
> > On Thu, 12 Feb 2009 08:16:45 -0800 |
5 |
> > Donnie Berkholz <dberkholz@g.o> wrote: |
6 |
> > > I think we should update the PMS to bash 3.1 to allow for '+=' |
7 |
> > > use. I looked through the bash changelog, and 3.2 didn't appear |
8 |
> > > to add any new and useful features. We should then require a new |
9 |
> > > EAPI for new bash features. |
10 |
> > |
11 |
> > That would require GLEP 55 to provide any protection. A new EAPI on |
12 |
> > its own wouldn't be enough. |
13 |
> |
14 |
> What kind of protection are you talking about |
15 |
|
16 |
The problem is, without GLEP 55, EAPI isn't known before the ebuild is |
17 |
sourced to generate metadata. If someone uses += anywhere that older |
18 |
bash looks when sourcing for metadata generation (which is not just |
19 |
global scope), the package manager won't know that the EAPI says that |
20 |
bash-3.1 is required for sourcing until after it's already done the |
21 |
sourcing, by which point it's too late. |
22 |
|
23 |
> and is it already in place for every bash feature in 3.0? |
24 |
|
25 |
The bash 3.0 transition was done before EAPIs came along, and was |
26 |
handled by the old fashioned "wait for ages until we're absolutely sure |
27 |
that everyone has bash 3.0 before continuing" method. |
28 |
|
29 |
-- |
30 |
Ciaran McCreesh |