3.5 Snapshots (Again)

3.5 Snapshots (Again)

[Ranjit Mathew]

Hi,

    The last time this issue was raised by
Kelly Cook nothing happened, so I'm raising this
yet again: can we please start having 3.5 snapshots?

Now that we are able to publish snapshots from
two series thanks to the changes made by Gerald,
I think we should start publishing 3.4 and 3.5
snapshots instead of persisting with 3.3 and 3.4
snapshots.

The policy should be to publish snapshots from
the current development tree (mainline) and the
current release tree - as I said before, I don't
think there are hordes of hackers just waiting for
their weekly 3.3 snapshots at this point.

Thanks,
Ranjit.

PS: If/when tree-ssa gets merged, will the current
mainline aspire to become GCC 4.0.0?

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: 3.5 Snapshots (Again)

[Gerald Pfeifer]

On Mon, 8 Mar 2004, Ranjit Mathew wrote:
> The last time this issue was raised by Kelly Cook nothing happened, so
> I'm raising this yet again: can we please start having 3.5 snapshots?

First we need some infrastructure changes to get the diffs right,
that's the major blocker right now.

I'll try to work on that tonight.

> The policy should be to publish snapshots from the current development
> tree (mainline) and the current release tree - as I said before, I don't
> think there are hordes of hackers just waiting for their weekly 3.3
> snapshots at this point.

Snapshots are not just used by hacker, but also by the FreeBSD system
and probably other as well.  And there are quite few hackers which do
not have CVS access.  In fact, it is hardly possible to contribute code
to GCC with only snapshot access, as one cannot do proper testing.

Gerald

Re: 3.5 Snapshots (Again)

[Ranjit Mathew]

Gerald Pfeifer wrote:
>>The last time this issue was raised by Kelly Cook nothing happened, so
>>I'm raising this yet again: can we please start having 3.5 snapshots?
> 
> 
> First we need some infrastructure changes to get the diffs right,
> that's the major blocker right now.

If you mean the diffs from the last 3.4 snapshot
to the newly generated 3.5 snapshot so that
people with current snapshots can transition
easily, yes I agree. But then, I think these diffs
are going to be pretty big anyway and it might be
better to avoid them altogether and just download
the whole snapshot once more (and continue
with diffs in subsequent weeks).


> I'll try to work on that tonight.

Thank you.


>>The policy should be to publish snapshots from the current development
>>tree (mainline) and the current release tree - as I said before, I don't
>>think there are hordes of hackers just waiting for their weekly 3.3
>>snapshots at this point.
> 
> 
> Snapshots are not just used by hacker, but also by the FreeBSD system
> and probably other as well.  And there are quite few hackers which do
> not have CVS access.  In fact, it is hardly possible to contribute code
> to GCC with only snapshot access, as one cannot do proper testing.

I didn't know about the FreeBSD part...

I am one of those hackers who do not have CVS access,
mainly because of bandwidth problems (dial-up is quite
unreliable here and "broadband" is expensive and
unnecessarily restrictive).

I tend to test and submit patches, if any, close
to the release of the snapshots to avoid being too
out of sync with CVS. In any case, my changes are
mostly to GCJ which doesn't violently change every
week. :-)

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: 3.5 Snapshots (Again)

[Gerald Pfeifer]

On Tue, 9 Mar 2004, Ranjit Mathew wrote:
>> First we need some infrastructure changes to get the diffs right,
>> that's the major blocker right now.
> If you mean the diffs from the last 3.4 snapshot to the newly generated
> 3.5 snapshot so that people with current snapshots can transition
> easily, yes I agree. But then, I think these diffs are going to be
> pretty big anyway and it might be better to avoid them altogether and
> just download the whole snapshot once more (and continue with diffs in
> subsequent weeks).

Well, if I don't have a good solution for the diffs, that's what I'll
do anyways.

I have just improved the gcc_release script a bit, and will test it
tomorrow and then use that for Monday's snapshot.  Then we can give
a 3.5 snapshot with diffs a try.

Gerald