F. Bluemers wrote:I cannot reproduce #2 at the moment,I always have at least 1 pv.
Are there any special rybka settings you are using?
how big did you set multipv?
Best
Fonzy
Nothing special that I'm aware of. I've been using MultiPV=4. It's most likely OS related, since I'm still using Window98. Unless someone else has this problem while using a newer OS you should probably just ignore it.
Ok,probably.Just curious ,do other engines behave the same?
btw fixing #1 and #3 will be more complicated.
I'll look into it.hope this fix will be usable till then
Best
Fonzy
I just tried Glaurung 2.1 and it's about the same. At shallow depths it often will just add each new variation as it finds it, but when it gets deeper (say, over a minute for a ply) it usually displays one line and quickly erases it, then displays two lines and quickly erases it, etc. until it show them all, which stay up until it has finished the first variation at the next depth.
Ok,In the next couple of days i'll dig up my (t)rusty win98se cdrom
and do some experiments.
Best
Fonzy
This generates NULL pv strings when compiled with VS2003, probably any _MSC_VER compiler.
In adapter.cpp, '%lld' needs to be replaced with '" S64_FORMAT "' in several places to work correctly because for an _MSC_VER compiler, node_nb will be of type signed __int64, and require '%I64d' instead of '%lld' for the format.
What happens is the vprintf() library function doesn't calculate the size of node_nb correctly, and then hammers following arguments, which are frequently pv strings.
I wanted a win95 compatible binary, and after removing the affinity stuff, discovered this problem. VS2003 is the last version which supports win95 binaries, so that's why I was using VS2003.
bnemias wrote:This generates NULL pv strings when compiled with VS2003, probably any _MSC_VER compiler.
In adapter.cpp, '%lld' needs to be replaced with '" S64_FORMAT "' in several places to work correctly because for an _MSC_VER compiler, node_nb will be of type signed __int64, and require '%I64d' instead of '%lld' for the format.
What happens is the vprintf() library function doesn't calculate the size of node_nb correctly, and then hammers following arguments, which are frequently pv strings.
I wanted a win95 compatible binary, and after removing the affinity stuff, discovered this problem. VS2003 is the last version which supports win95 binaries, so that's why I was using VS2003.
Hi Bob
Thanks for the bugreport.I'll fix 'adapter.ccp'
As for microsoft:
I compiled pg with vc c++ 2005 express.
No problems printing pv strings.
I tested the compiled exe under windows 98se with virtual pc and it seems to works well.