I've discovered what seems to be a workaround for the buffer overflow problem I described in my June 23, 2020, message to this list. This might hint at the source of the problem (or it might mean that the problem is now moot).
The issue was that when analyzing certain games using Stockfish version 230620 64 BMI2 in Scid version 4.6.4, typically after stopping and later restarting the analysis, Scid would analyze two moves and then crash, with a command-line error message "buffer overflow detected: /usr/games/tkscid terminated Aborted." The same problem occurred with later versions of Scid (version 4.6.4 is the current one in the Debian repositories) as well as with later versions of Stockfish. It did not occur with Scid Vs. PC; there was also no problem when analyzing individual positions from the game using Stockfish from the command line. One possibly relevant point that I had not mentioned before: when running long analyses with Scid, the program "top" tends to report unusually high memory usage (e.g. if 50% of the machine's memory is allocated for hash, "top" reports that the analysis starts at 50% or 51% of RAM and then slowly creeps up to sometimes 80% or more). It is possible that this has to do with quirks of how "top" reports memory usage rather than some memory leak. The way I seem to have been able to circumvent the buffer overflow problem is to create a separate PGN file where the starting position, indicated with a FEN string in the PGN file headers, is the position just before the last move that had been successfully annotated by Scid, and then to insert the rest of the moves from the game into this separate PGN file. This new PGN file can then be opened in Scid and analyzed. Then one can manually combine the analyzed portion of the first PGN file with the analysis in the second PGN file. So, seemingly, the buffer overflow problem is not caused by some feature of the position being analyzed, but has something to do with whatever Scid is doing with the previous moves contained in the PGN file. Also possibly relevant is that the buffer overflow crash that occurs when restarting analysis in a game that had previously been analyzed up to a certain move also occurs (and also after two moves) when restarting the analysis from any later position (not just the last analyzed move). Regards, Greg Marks
signature.asc
Description: PGP signature
_______________________________________________ Scid-users mailing list Scid-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scid-users
