Turo Lamminen

http://valgrind.org/docs/manual/mc-manual.html#mc-manual.mempools

Stack trace with memory pool disabled: ``` Invalid read of size 1 at 0x158B48: R_DrawColumn (r_draw.c:139) by 0x15B1DD: R_RenderSegLoop (r_segs.c:336) by 0x15B796: R_StoreWallRange (r_segs.c:711) by 0x166BBA: R_ClipPassWallSegment (r_bsp.c:218) by 0x166F8B:...

The WADs BLOCKMAP might be broken. One frame up the stack: ``` (gdb) insp *list $1 = -257 ``` changing line 489 to ``` for ( list = blockmaplump+offset ;...

Most likely vanilla just reads some random memory that just happens to not crash.

I'm not convinced that `continue` is the right thing to do. What happens if there's no -1 in the whole BLOCKMAP? There should also be a range check which terminates...

Looks OK other than the code style. Do you want to rebase/squash this into one commit or should it be done when merging?

There was a PR that made the textscreen and main game share the same renderer. How will that be affected by this change or always forcing software renderer for textscreen?

Check `HACKING.md` for coding style documentation. If you rebase this on latest `master` there should now be a linting CI check that should comment on any style violations.

For E2M1 Valgrind says: ``` Conditional jump or move depends on uninitialised value(s) at 0x14AB8B: P_BlockLinesIterator (p_maputl.c:488) by 0x149A53: P_CheckPosition (p_map.c:465) by 0x149A53: P_CheckPosition (p_map.c:402) by 0x149AA5: P_TryMove (p_map.c:490) by...

Are you using the latest master or some release build? This should have been fixed already. If it still happens with latest master, post contents of stdout.