At 12:27 PM 6/28/2005, Mladen Turk wrote: >William A. Rowe, Jr. wrote: > >> Question; do we want to adopt (finish adopting) svn's build >>system coded in python? > >Are you guys saying that I would need a python to even >consider to build a APR from the cvs/svn?
This is true *today* for everyone except Win32 (and netware?) >IMHO that would be a huge step backward. It's bit me. I'm unlikely to be contributing much on HP/UX or AIX for the foreseeable future until I reengineer my entire build environment. BSD is not much better, I can't even do simple scratch tests on the people.apache.org box. Yes we are *strictly* talking about running autoconf; once the autoconf is done, ./configure shouldn't require Python (right?) But alot of APR users want to live on the bleeding edge, and we want to encourage folks to make improvements/submit them back, which implies they live and breathe the svn trunk. >First of all, the number of files in the APR will unlikely >go sky-high, so what's wrong with the current build? >Seems to me that we are over-engineering here. <nods> I am trying to discover what the python solution actually solves. Earlier versions all worked quite lovely imho, although already quite deep in AC/libtool version dependencies. We seem to be taking large steps backwards - we are trying to craft the most portable implementation layer that is efficient on a wide range of platforms, yet the label gnu/apr is starting to swim in my head :) I can't decide which is more efficient; pull us further in this direction by bringing in the svn solution for win32 (which would require python to ./configure, IIUC), or look to our OpenSSL sister group for their perl build system with much fewer dependencies and more friendly build tool licenses. Bill
