On 08/03/2010 03:05 PM, Georg S. Weber wrote:
> On 3 Aug., 08:34, Rob Beezer <[email protected]> wrote:
>> I just built a clone in about a minute.  On a fresh install of
>> 4.5.2.rc0 created from source.  Maybe with a binary install, the docs
>> need to be built on the first clone?
> 
> No,
> 
> the problem is with relocation, but not with binary installs "per se".
> Proof: take a new (working ...) source tarball, build Sage by issuing
> "make". Issuing "make" a second time finishes without doing some real
> work. Then move this Sage installation to another position in the
> directory tree (probably you just need to rename the dir that
> $SAGE_ROOT is). This is the important step, the relocation step! Now,
> issue "make" again --- and whooops, all of the just newly built docs
> are considered "dirty" and are going to be built again.
> 
> This of course hurts almost any binary install, since they virtually
> never are installed in exactly the same place in the directory tree,
> where they were built ...
> (I'd bet that if one were installing a binary install in *exactly* the
> same location it was originally built, e.g. on a different computer,
> then sage-clone would be fast, i.e. would not think it had to rebuild
> the docs!)
> 
> So the problem lies buried in how the "history/memory" in Sphinx
> works, i.e. how Sphinx marks its generated files as dirty or not. I
> fear some type of pickling is done, where the full paths of the
> documentation files are used. But many people have spent already a
> substantial amount of time with Sphinx issues, so I am reluctant to
> even take a closer look.

By the way, making "make doc" relocation-safe is

http://trac.sagemath.org/sage_trac/ticket/8258

Your idea to use content hashes instead of timestamps seems very
promising.  Unfortunately, I haven't had the time to implement it in
full and try to push the proposed changes upstream.

I just noticed that John Palmieri has inquired about relocation on
sphinx-dev:

http://groups.google.com/group/sphinx-dev/browse_thread/thread/aa2c41b4708c726c#

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to