[gentoo-pms] Bash features

[gentoo-pms] Bash features

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 324 bytes --]

Hi,

as I now learnt, Bash 3.2 vanilla is not enough for some eclasses to
run.  Patchlevel 48 is needed.  Should we tighten the version string
for Bash?

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 394 bytes --]

On Sat, 9 Jan 2010 00:42:01 +0100
Christian Faulhammer <fauli@gentoo.org> wrote:
> as I now learnt, Bash 3.2 vanilla is not enough for some eclasses to
> run.  Patchlevel 48 is needed.  Should we tighten the version string
> for Bash?

Didn't the Council say that if it ever happened again, there should be
forced reverts rather than updating PMS retroactively?

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[Ulrich Mueller]

>>>>> On Sat, 9 Jan 2010, Ciaran McCreesh wrote:

> On Sat, 9 Jan 2010 00:42:01 +0100
> Christian Faulhammer <fauli@gentoo.org> wrote:
>> as I now learnt, Bash 3.2 vanilla is not enough for some eclasses
>> to run. Patchlevel 48 is needed.

What eclasses are this? _p48 isn't marked as stable.

>> Should we tighten the version string for Bash?

I don't think there's a need for this, as the patches are only
bugfixes.

> Didn't the Council say that if it ever happened again, there should
> be forced reverts rather than updating PMS retroactively?

Yes, but any =bash-3.2* is still allowed:

| Vote (6 yes, 1 no): Ebuilds must be completely parsable with
| =bash-3.2*, any use of later bash features will be reverted.

See also the log [1] where the example of 3.2_p39 is explicitly
mentioned at one point.

Ulrich

[1] <http://www.gentoo.org/proj/en/council/meeting-logs/20091109.txt>

Re: [gentoo-pms] Bash features

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 672 bytes --]

On Sat, 9 Jan 2010 09:18:19 +0100
Ulrich Mueller <ulm@gentoo.org> wrote:
> > Didn't the Council say that if it ever happened again, there should
> > be forced reverts rather than updating PMS retroactively?
> 
> Yes, but any =bash-3.2* is still allowed:
> 
> | Vote (6 yes, 1 no): Ebuilds must be completely parsable with
> | =bash-3.2*, any use of later bash features will be reverted.

Doesn't the way that's worded mean "it has to work with every single
3.2 patch", though? I take that to mean "must be completely parsable
with bash-3.2_p0 through bash-3.2_p9999", not "must work with at least
one version of bash-3.2_psomething".

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[Petteri Räty]

[-- Attachment #1: Type: text/plain, Size: 786 bytes --]

On 01/09/2010 10:58 AM, Ciaran McCreesh wrote:
> On Sat, 9 Jan 2010 09:18:19 +0100
> Ulrich Mueller <ulm@gentoo.org> wrote:
>>> Didn't the Council say that if it ever happened again, there should
>>> be forced reverts rather than updating PMS retroactively?
>>
>> Yes, but any =bash-3.2* is still allowed:
>>
>> | Vote (6 yes, 1 no): Ebuilds must be completely parsable with
>> | =bash-3.2*, any use of later bash features will be reverted.
> 
> Doesn't the way that's worded mean "it has to work with every single
> 3.2 patch", though? I take that to mean "must be completely parsable
> with bash-3.2_p0 through bash-3.2_p9999", not "must work with at least
> one version of bash-3.2_psomething".
> 

At least my opinion is all instead of any.

Regards,
Petteri


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-pms] Bash features

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 1073 bytes --]

Hi,

Ciaran McCreesh <ciaran.mccreesh@googlemail.com>:

> On Sat, 9 Jan 2010 09:18:19 +0100
> Ulrich Mueller <ulm@gentoo.org> wrote:
> > > Didn't the Council say that if it ever happened again, there
> > > should be forced reverts rather than updating PMS retroactively?
> > 
> > Yes, but any =bash-3.2* is still allowed:
> > 
> > | Vote (6 yes, 1 no): Ebuilds must be completely parsable with
> > | =bash-3.2*, any use of later bash features will be reverted.
> 
> Doesn't the way that's worded mean "it has to work with every single
> 3.2 patch", though? I take that to mean "must be completely parsable
> with bash-3.2_p0 through bash-3.2_p9999", not "must work with at least
> one version of bash-3.2_psomething".

 So one has to work around obvious bugs to satisfy p0?  p9999 will
support everything p0 has while eliminating bugs that cripple official
features advertised for p0.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[Ulrich Mueller]

>>>>> On Sat, 9 Jan 2010, Ciaran McCreesh wrote:

>> | Vote (6 yes, 1 no): Ebuilds must be completely parsable with
>> | =bash-3.2*, any use of later bash features will be reverted.

> Doesn't the way that's worded mean "it has to work with every single
> 3.2 patch", though? I take that to mean "must be completely parsable
> with bash-3.2_p0 through bash-3.2_p9999", not "must work with at
> least one version of bash-3.2_psomething".

Then we could simply have said "must be parsable with =bash-3.2" which
we didn't. Clearly, the main idea was to disallow usage of bash-4*
features. We could also discuss if "use of features" includes things
mentioned in the documentation but not properly working because of
bugs. Anyway, I think that tweaking the spec's wording wouldn't have
many practical consequences.

But we may have to revert that eclass for a different reason:
The council has voted (in the same meeting) that an upgrade path for
stable systems must be provided for one year at least. One year ago
bash-3.2_p33 was the latest stable version, therefore we must not yet
rely on any bugfixes in _p48.

Ulrich

Re: [gentoo-pms] Bash features

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

Hi,

Ulrich Mueller <ulm@gentoo.org>:
> But we may have to revert that eclass for a different reason:
> The council has voted (in the same meeting) that an upgrade path for
> stable systems must be provided for one year at least. One year ago
> bash-3.2_p33 was the latest stable version, therefore we must not yet
> rely on any bugfixes in _p48.

 It may be any patch level that makes python.eclass workable.  I only
could test the difference between p0 and p48.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[David Leverton]

On Saturday 09 January 2010 18:44:02 Ulrich Mueller wrote:
> > Doesn't the way that's worded mean "it has to work with every single
> > 3.2 patch", though? I take that to mean "must be completely parsable
> > with bash-3.2_p0 through bash-3.2_p9999", not "must work with at
> > least one version of bash-3.2_psomething".
>
> Then we could simply have said "must be parsable with =bash-3.2" which
> we didn't. Clearly, the main idea was to disallow usage of bash-4*
> features.

That would completely defeat the purpose of specifying a version at all.

The most sensible interpretation would be to treat the bash version 
requirement in the same way as an ebuild dependency.  If an ebuild says 
>=app-shells/bash-3.2 (it should be >= here, not =*, because ebuilds need to 
work with bash 4 too, but the principle is the same), then it means the 
ebuild is expected to work with /any/ version that matches the dependency.  
The only difference is that we have a spec that defines the "dependency" 
which ebuilds are supposed to respect, as opposed to documenting the 
requirements of a package that someone already wrote.

Re: [gentoo-pms] Bash features

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 930 bytes --]

Hi,

David Leverton <levertond@googlemail.com>:
> The most sensible interpretation would be to treat the bash version 
> requirement in the same way as an ebuild dependency.  If an ebuild
> says 
> >=app-shells/bash-3.2 (it should be >= here, not =*, because ebuilds
> >need to 
> work with bash 4 too, but the principle is the same), then it means
> the ebuild is expected to work with /any/ version that matches the
> dependency. The only difference is that we have a spec that defines
> the "dependency" which ebuilds are supposed to respect, as opposed to
> documenting the requirements of a package that someone already wrote.

 We can leave the spec as it is, in the end it was a bug fix which
resulted in the behaviour we wanted to specify.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-pms] Bash features

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 1044 bytes --]

Hi,

Ulrich Mueller <ulm@gentoo.org>:

> >>>>> On Sat, 9 Jan 2010, Ciaran McCreesh wrote:
> 
> > On Sat, 9 Jan 2010 00:42:01 +0100
> > Christian Faulhammer <fauli@gentoo.org> wrote:
> >> as I now learnt, Bash 3.2 vanilla is not enough for some eclasses
> >> to run. Patchlevel 48 is needed.
> 
> What eclasses are this? _p48 isn't marked as stable.

 So it is another patch level, but vanilla Bash 3.2 won't work with
python.eclass at different positions (mostly if conditions, can't tell
what part exactly).

> >> Should we tighten the version string for Bash?
> 
> I don't think there's a need for this, as the patches are only
> bugfixes.

 Sure, but the tree is hit by a needed bug fix.  I researched a lot,
because I bootstrapped with vanilla 3.2 and did not accuse it of the
failures I saw.  Maybe something like "the latest available" is doable.

V-Li

-- 
Christian Faulhammer, Gentoo Lisp project
<URL:http://www.gentoo.org/proj/en/lisp/>, #gentoo-lisp on FreeNode

<URL:http://gentoo.faulhammer.org/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]