On 2016-11-22 18:35, Robert Bradshaw wrote:
What I'm not following here is how this is not the correct behavior,
nor how hashes would help other than adding complexity (e.g. one must
store the hash of a.pxd inside b.c). Also, manually setting timestamps
would risk incorrect builds (remember the s
OK, let's consider the following setup: we have three Cython modules,
a, b, and c, and b cimports a, so we have (at least) three files
a.pyx
a.pxd
b.pyx
c.pyx
Having just built sage, we also have
a.c
b.c
c.c
with *newer* timestamps than any of the above. Now we switch branches
(forward, backwar
On Fri, Nov 18, 2016 at 8:18 AM, William Stein wrote:
>
> Ideas:
>
>
> What about changing Cython to optionally use sha1 or md5 hashes instead of
> timestamps?
I've had this thought as well--it would be fairly straightforward to
output a list of hashes for all Cython sources and compare those whe
I think thats backwards, you want to control timestamps and the answer is
to script some convoluted way of merging branches for very specific
circumstances? Why not control timestamps?
You really need to a) save timestamps and b) restore timestamps for
unmodified files. To be reliable a) needs
It would be great is at least the developer docs get an short explanation
of how to do this with plain git.
(or at least *here* :-))
Deja vu for me: I
posted https://groups.google.com/d/msg/sage-devel/hea2Ncbki7Y/cV47AWjDBwAJ
a while ago, and now I don't even remember how exactly 'git workspace'
So, back to the git worktree plan? I had it mostly working before
switching to this approach. If $SAGE_ROOT is /path/to/sage-7.4, it checks
out to a sibling folder /path/to/sage-7.4_merge_tree. Automating it is a
bit more complicated though; if the merge fails and the user has to correct
it manu
Thats a wrong-way merge and is confusing if one ever wants to understand
the history of the ticket. The convention is that the feature branch is
first, and you merge in dependencies.
On Sunday, November 20, 2016 at 2:31:57 AM UTC+1, David Roe wrote:
>
> I've implemented a new git-trac command
I've implemented a new git-trac command at
https://github.com/sagemath/git-trac-command/pull/26. I ended up not using
git worktree at all. Instead, the command creates a new branch from
develop, and merges in the changes from the old branch. I know that
there's been some discussion that merging
On Sat, Nov 19, 2016 at 7:12 AM, Jeroen Demeyer
wrote:
> On 2016-11-19 11:23, David Roe wrote:
>
>> I mean that even if I checkout develop again before building Sage, it
>> still takes a long time.
>>
>
> You mean something like
>
> - checkout old branch (but don't build it)
> - checkout new bran
On 2016-11-19 11:23, David Roe wrote:
I mean that even if I checkout develop again before building Sage, it
still takes a long time.
You mean something like
- checkout old branch (but don't build it)
- checkout new branch
- build Sage
That's a known issue, but it's certainly not Sage-specific
I mean that even if I checkout develop again before building Sage, it still
takes a long time.
David
On Sat, Nov 19, 2016 at 3:35 AM, Jeroen Demeyer
wrote:
> On 2016-11-18 08:12, David Roe wrote:
>
>> If I checkout an old branch (say, from one or two versions of Sage ago),
>> it essentially forc
On 2016-11-18 08:12, David Roe wrote:
If I checkout an old branch (say, from one or two versions of Sage ago),
it essentially forces a rebuild of all of Sage, even if I think better
of it and checkout develop immediately.
What exactly do you mean with "even if I think better
of it and checkout
Ideas:
What about changing Cython to optionally use sha1 or md5 hashes instead of
timestamps?
Or write a python script that just sets the time stamps back on all the
files that haven't changed. (Run it before and after...).
On Thu, Nov 17, 2016 at 11:12 PM David Roe wrote:
> If I checkout an
If I checkout an old branch (say, from one or two versions of Sage ago), it
essentially forces a rebuild of all of Sage, even if I think better of it
and checkout develop immediately. The rebuild is a consequence of the fact
that Cython builds based on timestamp and all of the files have been
touc
14 matches
Mail list logo