HPPA and Erlang packages

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

HPPA and Erlang packages

Sergei Golovan
Hi release managers!

I'd like to ask you about what to do with Erlang and its reverse
dependencies on hppa architecture. The problem is that there's a bug
with fork()+exec() which makes erlang FTBFS (and the currently built
packages are broken as well) on hppa (see [1], [2]). It seems to be
very complicated (as it's known and hasn't been fixed for about a
three months) and prevents new Erlang packages migration to testing.

So, what to do with Erlang in testing and unstable? I'd prefer to
remove erlang and its reverse dependencies from both testing and
unstable for hppa architecture (all the packages don't work anyway,
and nobody uses them, given that there's no bugreports). After the bug
will be fixed, all the packages will be eventually rebuilt.

[1] http://lists.debian.org/debian-hppa/2009/12/msg00035.html
[2] http://article.gmane.org/gmane.linux.ports.parisc/2403

Cheers!
--
Sergei Golovan


--
To UNSUBSCRIBE, email to [hidden email]
with a subject of "unsubscribe". Trouble? Contact [hidden email]
Archive: http://lists.debian.org/f60b7eb61002190427g2aee0658p685d94ace2462fd3@...

Reply | Threaded
Open this post in threaded view
|

Re: HPPA and Erlang packages

Marc 'HE' Brockschmidt-3
Sergei Golovan <[hidden email]> writes:
> I'd like to ask you about what to do with Erlang and its reverse
> dependencies on hppa architecture. The problem is that there's a bug
> with fork()+exec() which makes erlang FTBFS (and the currently built
> packages are broken as well) on hppa (see [1], [2]). It seems to be
> very complicated (as it's known and hasn't been fixed for about a
> three months) and prevents new Erlang packages migration to testing.

Dear HPPA porters, what's the status here?

> So, what to do with Erlang in testing and unstable? I'd prefer to
> remove erlang and its reverse dependencies from both testing and
> unstable for hppa architecture (all the packages don't work anyway,
> and nobody uses them, given that there's no bugreports).

Yes, the binary removal is the best option, if the porters are not able
to fix this arch-specific problem. I would like to avoid this at all
costs, but it might be the only available solution.

Carlos, could you, as HPPA porter, shed some light on this? Can we
expect this to be fixed in the next few weeks? Is anyone actually still
trying to fix this?

Marc
--
BOFH #375:
Root name servers corrupted.

attachment0 (202 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: HPPA and Erlang packages

Sergei Golovan
On Mon, Feb 22, 2010 at 8:43 PM,  <[hidden email]> wrote:
>
> Yes, the binary removal is the best option, if the porters are not able
> to fix this arch-specific problem. I would like to avoid this at all
> costs, but it might be the only available solution.

I have added a workaround which disables using vfork() on hppa
architecture (it is substituted with fork()), so, there's no need in
removing binary packages. Though it would be better to fix vfork()
anyway.

Cheers!
--
Sergei Golovan


--
To UNSUBSCRIBE, email to [hidden email]
with a subject of "unsubscribe". Trouble? Contact [hidden email]
Archive: http://lists.debian.org/f60b7eb61002221003l32e878b7ma2e13f408fc0ef53@...

Reply | Threaded
Open this post in threaded view
|

Re: HPPA and Erlang packages

Carlos O'Donell-2
On Mon, Feb 22, 2010 at 1:03 PM, Sergei Golovan <[hidden email]> wrote:

> On Mon, Feb 22, 2010 at 8:43 PM,  <[hidden email]> wrote:
>>
>> Yes, the binary removal is the best option, if the porters are not able
>> to fix this arch-specific problem. I would like to avoid this at all
>> costs, but it might be the only available solution.
>
> I have added a workaround which disables using vfork() on hppa
> architecture (it is substituted with fork()), so, there's no need in
> removing binary packages. Though it would be better to fix vfork()
> anyway.

Sergei,

This is the workaround I'm going to suggest for now.

I am still working on this issue. Unfortunately I've been pulled away
to track down an HPPA kernel boot issue involving PCI cache line
sizes, and thankfully this issue was just fixed yesterday.

I am back again looking at why vfork() on hppa is failing. I don't
have a good estimate for how long this will take, given that we don't
really understand the nature of the failure. I will keep debian-hppa
updated with the status of my investigations into the issue.

Cheers,
Carlos.


--
To UNSUBSCRIBE, email to [hidden email]
with a subject of "unsubscribe". Trouble? Contact [hidden email]
Archive: http://lists.debian.org/119aab441002230502n22d558e9vc54d51503078ccff@...

Reply | Threaded
Open this post in threaded view
|

Re: HPPA and Erlang packages

Marc Brockschmidt-4
"Carlos O'Donell" <[hidden email]> writes:

> On Mon, Feb 22, 2010 at 1:03 PM, Sergei Golovan <[hidden email]> wrote:
>> On Mon, Feb 22, 2010 at 8:43 PM,  <[hidden email]> wrote:
>>> Yes, the binary removal is the best option, if the porters are not able
>>> to fix this arch-specific problem. I would like to avoid this at all
>>> costs, but it might be the only available solution.
>> I have added a workaround which disables using vfork() on hppa
>> architecture (it is substituted with fork()), so, there's no need in
>> removing binary packages. Though it would be better to fix vfork()
>> anyway.
> This is the workaround I'm going to suggest for now.
Great, this allows us to ignore the problem for some more time.

> I am back again looking at why vfork() on hppa is failing. I don't
> have a good estimate for how long this will take, given that we don't
> really understand the nature of the failure. I will keep debian-hppa
> updated with the status of my investigations into the issue.

Good to know. Thanks for your work :-)

Marc
--
BOFH #152:
My pony-tail hit the on/off switch on the power strip.

attachment0 (203 bytes) Download Attachment