Sage git master?

[Jeroen Demeyer]

Sorry to ask this silly question, but what exactly is the "correct" git
master to use?

The page http://trac.sagemath.org/wiki/QuickStartSageGit points to
ssh://g...@trac.sagemath.org:2222/sage.git
but there is also
https://github.com/sagemath/sage
of which I was told once that "master" isn't actually the right branch,
it's "build_system".

This is all confusing to me...

[Nathann Cohen]

Hellooooooooo !

I am confused too. I used to use "build_system", then was told that it
was wrong. I now use the master from the page you use, but I would be
glad if it was somehow confirmed ^^;

Nathann

[Volker Braun]

public/sage-git/master is the integration branch, in particular it has working doctests and dev scripts. If you just want to build sage-git then this is the way to go. 

The commits in the integration branch are bound to become (part of) the sage-git tree, so you can start branching off public/sage-git/master if you want to work on anything that is unrelated to the git transition.

If you want to contribute to the git transition then you should work on the branch in the appropriate sub-ticket of  the git transition meta-ticket at #13015, and not on public/sage-git/master (which is the branch attached to #13015). These sub-tickets should never merge in the integration branch.

[Volker Braun]

I've updated the quick start guide.

On Wednesday, October 2, 2013 11:21:01 AM UTC+1, Jeroen Demeyer wrote:

[R. Andrew Ohana]

For what its worth, the main reason why the build_system branch hasn't been merged into "master" is that it is still up for review -- it is done (unlike the development scripts).

As Volker said, if you are trying to actually do work using the git repository you should base your work off of public/sage-git/master (so long as your work is not essential to the git transition).

--
You received this message because you are subscribed to the Google Groups "sage-git" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-git+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--

Andrew

[Robert Bradshaw]

> On Wednesday, October 2, 2013 11:21:01 AM UTC+1, Jeroen Demeyer wrote:
>>
>> Sorry to ask this silly question, but what exactly is the "correct" git
>> master to use?

Not silly at all, it's quite unclear.

> public/sage-git/master is the integration branch, in particular it has
> working doctests and dev scripts.

Is this always the case now? I've had mixed luck (mostly bad) with
this branch always passing doctests.

On Wed, Oct 2, 2013 at 2:20 PM, R. Andrew Ohana <andrew...@gmail.com> wrote:
> For what its worth, the main reason why the build_system branch hasn't been
> merged into "master" is that it is still up for review -- it is done (unlike
> the development scripts).

I'm not sure I follow this logic. Isn't build_system in
public/sage-git/master? Why can't master contain unreviewed code until
it has become *the* official release branch.

Why not make the github.com/sagemath/sage/master =
public/sage-git/master = the thing that most people should build off
of (assuming we can keep it stable). We can have
github.com/sagemath/sage/build_system for just the build_system
additions, and github.com/sagemath/sage/upstream for the raw,
converted sage releases. Then we wouldn't have to even have wikis and
forum threads detailing what one should use (which are usually
consulted only after cloning from master and building fails
miserably).

Of course even better would be to get out of this hybrid development model soon.

- Robert

[R. Andrew Ohana]

On Wed, Oct 2, 2013 at 2:40 PM, Robert Bradshaw <robe...@gmail.com> wrote:

> On Wednesday, October 2, 2013 11:21:01 AM UTC+1, Jeroen Demeyer wrote:
>>
>> Sorry to ask this silly question, but what exactly is the "correct" git
>> master to use?

Not silly at all, it's quite unclear.

> public/sage-git/master is the integration branch, in particular it has
> working doctests and dev scripts.

Is this always the case now? I've had mixed luck (mostly bad) with
this branch always passing doctests.

I think the biggest problem here is that too many changes to the dev scripts are pushed without correcting the doctests to reflect changed output. That said, the dev scripts still seem to be under very active development, so I don't think I would call them "stable" yet.

On Wed, Oct 2, 2013 at 2:20 PM, R. Andrew Ohana <andrew...@gmail.com> wrote:
> For what its worth, the main reason why the build_system branch hasn't been
> merged into "master" is that it is still up for review -- it is done (unlike
> the development scripts).

I'm not sure I follow this logic. Isn't build_system in
public/sage-git/master? Why can't master contain unreviewed code until
it has become *the* official release branch.

There are 2 main reasons why I've kept master separate (because I did not used to):

1) It is much easier to merge the releases that Jeroen's releases into a branch that only has no modifications from the exported repositories. (This can be done locally, but I see no reason to keep this branch private)

2) It gives a base on which to review the changes that have been made for the git repository.

Why not make the github.com/sagemath/sage/master =
public/sage-git/master = the thing that most people should build off
of (assuming we can keep it stable). We can have
github.com/sagemath/sage/build_system for just the build_system
additions, and github.com/sagemath/sage/upstream for the raw,
converted sage releases. Then we wouldn't have to even have wikis and
forum threads detailing what one should use (which are usually
consulted only after cloning from master and building fails
miserably).

This is an easy change to make.

Of course even better would be to get out of this hybrid development model soon.

- Robert

--

Andrew

[Robert Bradshaw]

Yes, these are certainly worth keeping around.

>> Why not make the github.com/sagemath/sage/master =
>> public/sage-git/master = the thing that most people should build off
>> of (assuming we can keep it stable). We can have
>> github.com/sagemath/sage/build_system for just the build_system
>> additions, and github.com/sagemath/sage/upstream for the raw,
>> converted sage releases. Then we wouldn't have to even have wikis and
>> forum threads detailing what one should use (which are usually
>> consulted only after cloning from master and building fails
>> miserably).
>
> This is an easy change to make.

Excellent.

If no one's opposed, I'd be happy to do that myself. We should
probably only push to master when all doctests pass, even if it lags
the active development on the dev scripts a bit. Is there a "known
all-tests-passing" version of the dev scripts, or should I do a binary
search to find one (unless I get lucky and the current tip passes)?

- Robert

[R. Andrew Ohana]

I did a little work on them yesterday, and fixed all doctest errors then -- although since then quite a number of doctests have changed due to a change in the default log level (testing them out right now). It seems like not all doctests in the dev scripts are properly cleaning up the temporary files/directories they create, so you should be a little careful with that for the moment.

- Robert

--

Andrew

[Volker Braun]

+1 for making the "master" branch the current integration branch. In particular, checking out the github master really should give you something working.

As far as I am concerned, the current public/sage-git/master can be used (and had some testing at SD53 last week).

I've made some changes to the dev scripts, but they are now passing doctests for me. We can probably update public/sage-git/master soon.

[R. Andrew Ohana]

I'll make the change

master -> upstream

build_system + dev_scripts (that I've verified to work) -> master

when I update the git repo to 5.12.rc1 later tonight.

--

You received this message because you are subscribed to the Google Groups "sage-git" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-git+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--

Andrew

[Simon King]

Hi Andrew,

On 2013-10-05, R. Andrew Ohana <andrew...@gmail.com> wrote:
> I'll make the change
>
> master -> upstream
> build_system + dev_scripts (that I've verified to work) -> master

Are you talking about github or trac?

And if it is the latter, what does this mean for development work? I.e.,
how to change from my current playground (public/sage-git/master on
trac) to the new one?

Best regards,
Simon

[R. Andrew Ohana]

On Sat, Oct 5, 2013 at 12:33 AM, Simon King <simon...@uni-jena.de> wrote:

Hi Andrew,

On 2013-10-05, R. Andrew Ohana <andrew...@gmail.com> wrote:
> I'll make the change
>
> master -> upstream
> build_system + dev_scripts (that I've verified to work) -> master

Are you talking about github or trac?

Both.

And if it is the latter, what does this mean for development work? I.e.,
how to change from my current playground (public/sage-git/master on
trac) to the new one?

It should be as simple as switching from using public/sage-git/master to master.
 

Best regards,
Simon

--
You received this message because you are subscribed to the Google Groups "sage-git" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-git+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--

Andrew

[Volker Braun]

Thanks!

Can you also merge the positively-reviewed dependencies in http://trac.sagemath.org/13015

I realize that its not much yet, but I'm confident that we can get more stuff ready soon.

[Robert Bradshaw]

Excellent, thanks!

[Robert Bradshaw]

I still have tests failing on a fresh build of that branch: http://patchbot.sagemath.org/log/0/Ubuntu/12.10/x86_64/3.5.0-17-generic/pi/2013-10-07%2002:15:05%20-0700

[Volker Braun]

It seems both are due to different interleaving of stdout and python's own output stream. They work for me. In any case, its pretty clear what to do. Will try to fix.

Speaking of the patchbot, whats the status there? I tried the github verson of the patchbot recently but that only tested one ticket...

[Robert Bradshaw]

On Mon, Oct 7, 2013 at 10:43 AM, Volker Braun <vbrau...@gmail.com> wrote:
> It seems both are due to different interleaving of stdout and python's own
> output stream. They work for me. In any case, its pretty clear what to do.
> Will try to fix.

Good.

> Speaking of the patchbot, whats the status there? I tried the github verson
> of the patchbot recently but that only tested one ticket...

It should just work. What did it do after testing that one ticket?