On 2010 Jan 2, at 23:49, Bisz István wrote: > Hi Viktor, > >> I'd like to kindly ask you to elaborate more. > > With the previous, default build, way of the Heap managements, the HVM for > MinGW tried to replace the CRT functions with its own malloc, free, etc. > This is a completely wrong solution in this context, is too weak.
I hope Przemek can comment on it. I still can't see the real reason here. We do it for some compilers and we don't do it for others. For MSVC, without this trick it gave memory errors on exit so it was clearly apparent. Maybe MinGW was the same, but it didn't report it? >> How about the other change? > > ---------------------------------------- > Total memory allocated: 2251241 bytes (29215 block(s)) > Warning, memory allocated but not released: 0 bytes (0 block(s)) Currently above msg will also appear when not compiled with fmstat. I'll commit a fix. > It is the message (with HB_USER_CFLAGS=-DHB_FM_STATISTICS), when the all > allocated blocks are released perfectly, if it is not adequate please change > it. > > Note: Is the ending message now for hbide and demoxbp. There are still two HBQT TOFIXes where leak is only prevented by explicit .prg calls. Plus we're not using the official C++ allocation override for HBQT yet, which may be a problem. What do you think? Brgds, Viktor _______________________________________________ Harbour mailing list (attachment size limit: 40KB) [email protected] http://lists.harbour-project.org/mailman/listinfo/harbour
