1-Jan-2003 20:29 Thomas E. Dickey wrote: > On Thu, 2 Jan 2003, Leonid Pauzner wrote:
>> 1-Jan-2003 18:37 Thomas Dickey wrote: >> > I put a snapshot of my current patch on >> > ftp://invisible-island.net/temp/ >> >> > which includes LP's HTLine change, modified so it seems to work properly. >> >> This is still an error, Tom. >> Your variant will use (MAX_LINE+epsilon)*(lines in the document) memory, >> and this is because the last line is allocated in pool in split_line(). >> The last (=unfinished) line should be handled differently >> then all the previous (=finished) lines. > I did run the program with find-leaks and with valgrind - neither showed > me a problem. There might be some scenario where the program leaks - we > can test for that. (Actually there were leaks, not related, in the > HTParse.c code iirc). This variant did not leak (comparing to dev.11) - it allocates *too much memory* in the pool, and pool is freed later - in HText_free. Perhaps, you did not see memory corruption since memory overrun within the same pool chunk is not seen by memory debugger. There may be another palliative variant using freeHTLine: we malloc/free the last line each time, but other lines are allocated in the pool. (To free the last line we store its pointer in other place). May supply a patch. >> Please undo this code to dev.12 state again. >> >> > I'm running short of time, will resume working on Lynx tomorrow. ; To UNSUBSCRIBE: Send "unsubscribe lynx-dev" to [EMAIL PROTECTED]
