Re: status update
Neal Norwitz <[email protected]> Tue, 28 Jan 2003 20:40:29 -0500
| Newsgroups | gmane.comp.python.snake-farm.user |
|---|---|
| Message-ID | <[email protected]> |
On Wed, Jan 29, 2003 at 01:36:35AM +0100, Laura Creighton wrote: > <snip> > > My point was the snake-farm is doing its job. It is a nice benefit. > > But it would be even better if we could get it stable, so there's only > > green. > > Is this still a 'not enough people and not enough hours in the day' > problem, or could something else happen to expedite matters? I hope the rest of this message doesn't sound like I'm complaining, but it probably will. I'm thankful for the snake-farm, but it takes a lot to get it up to speed and to keep it running smoothly. The biggest help would be finding someone that can help with the AIX problem described a bit more below. There's a combination of issues. Certainly more people and more time would help. Builds take a long time. I suspect several issues going on with this: NFS; older machines, limited by memory and/or CPU; other processes running, etc. For reference, my 3 year old PC builds everything from scratch in about 2 minutes. It seems most of the faster machines in the farm take twice that time. There are also potentially many different configurations/changes necessary. For example, a debug build, build without optimization, use a different compiler, turn a single define off, etc. Each clean build multiplies the time it takes. The Solaris hanging problem took a great deal of time to diagnose and work on a patch. The problem turned out to be an an apparent bug in the Solaris 8 operating system. This is no one's fault, but it was time consuming. A bunch of the current problems are due to the compilers not working. I'm completely stuck on the AIX problem. It seems no one else has this problem. I don't know how to debug it further. Importing a dynamic module (actually loading the .so works), but as soon as the module (.so) calls back into the interpreter, python crashes. The function addresses look fine. HP-UX isn't working now, due to some similar problems and fixes to Solaris. But HP-UX should be close to working again. I believe we are making progress. However, since python is always changing, sometimes we take a step back for each step forward. I'm really not complaining about this. It takes a lot of work to get everything under control. I believe we are making progress. It's just slower than I had hoped. Neal