[mythtv] [mythtv-commits] mythtv commit: r25892 by robertm

[Robert McNamara]

On Wed, Sep 1, 2010 at 6:21 PM, buzz <davidbuzz at gmail.com> wrote:
> I recommend partially reverting or modifying this commit for the following
> reasons:
>
> * the original URL/s  are the primary and definitive location of the files ,
> so should be kept in the code and still available.   I'd suggest at the very
> least putting them in a comment or something.

Putting them in a comment would be acceptable to me, having to rush to
fix the busted build script a dozen-odd times per year is not.  We
just have way, way too little manpower looking at Windows to be able
to spend 90% of that weffort trying to make the build script work.
It's hard enough trying to keep Myth working on Windows.

> * When the specific version of a dependency that we use is removed from it's
> URL, it's because it's been replaced with a newer/better version, so if the
> script continues to use to use an older "cached" version it makes it harder
> to correct the script ( smaller patches are easier).

That may be absolutely true, but on the other hand, switching versions
from a known good to an unknown means a whole new round of
troubleshooting as issues arise.

> * It is harder ( impossible) to push any faults found with the dependancies
> back upstream to the author/s if we are using an old version.

We have never had the manpower to do this, either.  If we had a
thriving Myth Windows community then this might be feasible, but as it
stands maybe two of us devs and a half dozen users are doing anything
worthwhile in this space, and I just don't have time for it.  As I try
to do real, useful work in Windows, I'm already chasing build failures
every single time I run the script.  It takes what would be an
already-scant few hours of work I can put towards the Windows build
per week and turns them into minutes.  I just can't keep up this way.

> My suggested solution is this:
>
> * always check the original URL, using a http HEAD request, and if it's not
> available any more at that URL, throw a meaningful message describing the
> fact that you should see if a newer version is available.  ( see
> http://www.w3.org/Protocols/rfc2616/rfc2616-sec9.html section 9.4 )
>
> * after doing the above, if the file is still available AND it's at the
> google code repository, download it from the google repo ( since it's
> aparently faster).
> * if the file is NOT in the google repo, but still at the primary URL,
> download from the primary URL.  Optionally, if we wanted to be tricky, we
> could auto-upload it direct to the google repo at this time, if we wanted
> to.

The Windows build script is already insanely overcomplicated as it is,
and I just can't see myself doing this for all the reasons listed
above-- if we had a full time Windows Build script maintainer, sure,
but when two of us have to split time between trying to keep it
building, trying to keep the script in line, and in the last few
seconds left after that trying to improve windows features... well, if
we go back to the old way, that will be the end of my involvement with
the Windows port.  I don't have the time or the patience to spend 90%+
of my time poking at a bash script.

Robert