PDA

View Full Version : Seq versioning



purple
02-18-2005, 08:48 AM
Ok, I'm starting to get frustrated by people who are having problems with running an old version from precompression on postcompression streams and expecting it to work.

Does anyone mind if I start versioning each of my patches so people can tell directly from the console if they have the right version or not? Too many people are trying to use dev patches and don't know what they are doing. I can do 5.0.0.17-purple1 or whatnot probably. Is that ok? It means that whomever takes my patch into cvs (if that is appropriate) will need to redo the versioning in a more official way.

Unless I hear something complaining about it, I'm gonna start doing that.

bonkersbobcat
02-18-2005, 10:46 AM
Ok, I'm starting to get frustrated by people who are having problems with running an old version from precompression on postcompression streams and expecting it to work.

Does anyone mind if I start versioning each of my patches so people can tell directly from the console if they have the right version or not? Too many people are trying to use dev patches and don't know what they are doing. I can do 5.0.0.17-purple1 or whatnot probably. Is that ok? It means that whomever takes my patch into cvs (if that is appropriate) will need to redo the versioning in a more official way.

Unless I hear something complaining about it, I'm gonna start doing that.
That makes sense to me. I would recommend that any xxx-name are only generated off of a diff to xxx. By this I mean that 5.0.0.17-purple1 would be a patch that is off of the cvs version 5.0.0.17. If someone else adds a patch on top of 5.0.0.17-purple1 and they call it 5.0.0.17-joe1, the patch needs to apply to the cvs version 5.0.0.17 and not the patched version 5.0.0.17-purple1. (It can contain the same changes, but needs to always be based off of the base name.)

Or stated another way the convention is where you have xxx-yyy in a version name the -yyy is a patch of of xxx.

In the case of a stacked patch where a patch is not off of cvs, but rather off of some other patch base, it would be named xxx-yyy-zzz. For example if once purple creates a patch off of the cvs base and calls it 5.0.0.17-purple1 and after that I create a patch on top of that I would call it 5.0.0.17-purple1-bb1. If my patch was to be applied to cvs and happened to included the changes that purple did, I would call it 5.0.0.17-bb1.

Discussion?

Hopefully this won't really be that much of an issue, because I would like to get stuff into cvs pretty quickly. On that topic do we thing that the current purple patch is complete enough to go to 5.0.0.18?

BB

icyman
02-18-2005, 10:54 AM
Sounds like a plan.

purple
02-18-2005, 11:08 AM
BB, the current posted patch has 3 outstanding issues:
1) Some wierd segfault possibly with buff wear off messages
2) Compression fails when compiled 64bit
3) Fragment buffer overflows

Current CVS has problem #3 also.

I do have some slight changes in my local tree and I'm valgrinding so far to try to track down #1. I'll post another patch at some point. I got no clue what to do about #2 and #3 I haven't been able to reproduce locally in any way.

Spaz
02-18-2005, 11:41 AM
Not a dev, but I also think that clearer versioning is important.