On 2 May 2012 04:57, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> FWIW, I think only developers not packagers would really be taking such
> a hit. I assume we'd continue to ship prebuilt lexer output in
> tarballs, so there'd seldom be a reason for a packager to need to run
> the tool. Given the extremely slow rate of churn of the lexer, it might
> not be necessary for most developers to have the tool installed, either,
> if we were willing to put the derived file into git.
Incidentally, I had an unrelated conversation with someone (I think it
might have been Heikki) a while back, where it was suggested that Flex
and Bison could be run through web services. This might actually make
hacking Postgres on windows far easier, because the last time I tried
to do that the hard way, I gave up, suspecting that there must be some
kind of Winflex bug that selectively manifests itself - certainly, the
population of windows hackers is small enough that it could go
unnoticed for quite a while. Such an approach could be part of the
solution to this problem (although, incidentally, Quex maintains
visual studio support quite well, and even has graphical instructions
here: http://quex.sourceforge.net/doc/html/intro/visual_studio_trouble.html
).
It might be the case that some kind of virtualisation and/or
authentication (Postgres community account required) could make this
approach practical. It just isn't the path of least resistance right
now. Visual studio builds would be far easier if we did this, which
might encourage more hackers to venture into Windows land.
--
Peter Geoghegan http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training and Services
