Good news
Moderator: Moderators
-
Dr.InfernO
- Posts: 223
- Joined: 18 Nov 2005, 13:55
Would be too much work?If someone compiles it on it... it should work probably.
Anyway, it appears that somehow there occur sync errors.
Got some when I played linux 32 bit gcc 3.4.6 vs linux 32 bit gcc 4.0.3 when dgunning, sometimes just after 15 minutes or so. Also MSVC vs. linux errors out quite soon, and much map deformation / non-flat maps seem to cause errors too sometimes.
So I suspect either uninitialized memory reads which didn't matter with same compiler because they always gave the same value, or some math functions I forgot to wrap, or invalid float to int conversion, or some other weird problem.
Well, one can never expect things to work the first time.
We'll probably wait until we got the professional bug reports in and see if that might give us a clue.
Also I noticed spring repathed when switching between 32 and 64 bit builds so that may point to a bug too.
Anyway, it appears that somehow there occur sync errors.
Got some when I played linux 32 bit gcc 3.4.6 vs linux 32 bit gcc 4.0.3 when dgunning, sometimes just after 15 minutes or so. Also MSVC vs. linux errors out quite soon, and much map deformation / non-flat maps seem to cause errors too sometimes.
So I suspect either uninitialized memory reads which didn't matter with same compiler because they always gave the same value, or some math functions I forgot to wrap, or invalid float to int conversion, or some other weird problem.
Well, one can never expect things to work the first time.
We'll probably wait until we got the professional bug reports in and see if that might give us a clue.
Also I noticed spring repathed when switching between 32 and 64 bit builds so that may point to a bug too.
-
Dr.InfernO
- Posts: 223
- Joined: 18 Nov 2005, 13:55
Well, you should be able to read. This for example.Dr.InfernO wrote:please excuse my next (stupid?) question?
Is it hard to compile? do i need skills for it?
If you've got linux on it and a moderately recent gcc version (anything >= 3.4 should do), then it should be pretty straightforward.
It's in branches/tvo2. We could do a quick test but it isn't really necessary IMHO.
Progress:
I just tested two games with Zaphod, one 64 bit linux gcc 4.0.3 vs. 32 bit MinGW gcc 3.4.2 and one same binary on my side (linux) vs. MSVC. Linux was server. Both worked much better then trunk, and better then previous tests, but still sync errors. On MSVC the Y coordinate got out of sync while moving a unit on the MSVC client, that went away when the unit stopped moving. After some fighting it stayed, but XZME remained in sync. On MinGW it worked for approx. 5-10 minutes or something, then XZ errors started popping up, which didn't go away.
What can be concluded is that the fixes of yesterday evening worked (including fixing 64 compat), because it errored out much later then yesterday.
So UMRs (unititialized memory reads) may be the problem, but finding them is kinda hard. Also float to int casts may still be a problem, I'm discussing this with Nicolas. Further, it might be the case I missed a transcedental function somewhere... If none of that applies, then I'm out of clues for a while...
Progress:
I just tested two games with Zaphod, one 64 bit linux gcc 4.0.3 vs. 32 bit MinGW gcc 3.4.2 and one same binary on my side (linux) vs. MSVC. Linux was server. Both worked much better then trunk, and better then previous tests, but still sync errors. On MSVC the Y coordinate got out of sync while moving a unit on the MSVC client, that went away when the unit stopped moving. After some fighting it stayed, but XZME remained in sync. On MinGW it worked for approx. 5-10 minutes or something, then XZ errors started popping up, which didn't go away.
What can be concluded is that the fixes of yesterday evening worked (including fixing 64 compat), because it errored out much later then yesterday.
So UMRs (unititialized memory reads) may be the problem, but finding them is kinda hard. Also float to int casts may still be a problem, I'm discussing this with Nicolas. Further, it might be the case I missed a transcedental function somewhere... If none of that applies, then I'm out of clues for a while...
Sounds like this is getting off to a good start. Hopefully we can fix all of these errors and get it all synced up in no time. Looks like it might take a bit longer than I originally thought but hey its the right direction. 
Hopefully I can get the linux lobby done so that people can start playing with the linux ver!
Hopefully I can get the linux lobby done so that people can start playing with the linux ver!