On Fri, Apr 05, 2002 at 01:16:11PM -0500, Vincent Béron wrote: > Enrico Horn a écrit : > > > > Hi > > > > On Friday 05 April 2002 19:06, Andreas Mohr wrote: > > > On Fri, Apr 05, 2002 at 06:46:27PM +0200, Martin Lexa wrote: > > > > Hello! > > > > > > > > Function CreateFileA get this file path (just for example): > > > > > > > > C:\\Program Files\\3DO\\Heroes3\\DATAh3bitmap.lod > > > > > > > > As you can see the correct path should be: > > > > > > > > C:\\Program Files\\3DO\\Heroes3\\DATA\\h3bitmap.lod > > > > > > > > This lead to error message "Can't initialize resources.Possible disk > > > > problem." > > > > > > > > I see the problem in the function DOSFS_DoGetFullPathName(), which in > > > > the end removes the ending backslash from the path. The attached patch > > > > fixes it. > > > > > > > > Bye, Martin > > > > > > > > > > Yep, that might fix *your* problem, but don't you think that code snippet > > > might have been there for a purpose ? > > > > > To find out this purpose was the purpose of my mail "DOSFS_DoGetFullPathName" > > I sent to this list more than a week ago. Until now no real answer has been > > given. > > > Are you sure that simply completely removing code doesn't break anything > > > here in other cases ? > > I cant think of any situation where this might be needed. > > I may be wrong though. > > Yet so far I have no explanation for this piece of code and > > removing it seems to work fine. > > > > Enrico > > [EMAIL PROTECTED] > > I tracked it to a patch by Uwe Bonnes, around the 2000/04/28. It brought > the dos_fs.c revision from 1.46 to 1.47. The CVS comment is: > "DOSFS_DoGetFullPathName: rewrite to return resulta like OSR2". > > Have you tried to run HOMAM III under OSR2? Or even just a call to the > proper Win32 API? > It might be just some different behaviours under different Windows > falvours... > > As to which of the two is what we want, I can't help you on that. Actually GetFullPathName was known to have some trailing \\ issue. So this had to be fixed somehow. It just astonished me a bit to see some code part completely removed instead of a "real" fix. But he said that he can't think of any occasion where it breaks, so it might be ok.
Probably a candidate for regression testing, though ;-) -- Andreas Mohr Stauferstr. 6, D-71272 Renningen, Germany Tel. +49 7159 800604 http://home.arcor.de/andi.mohr/
