Carl Sorensen <[email protected]> writes: > On 5/27/10 2:58 PM, "Benjamin Peterson" <[email protected]> wrote: > >> Carl Sorensen <c_sorensen <at> byu.edu> writes: >>> I saw the comment from David not as a suggestion to build an automated >>> process for feeding the erroneous strings, but as a one-time testing >>> suggestion. Hence my request for help. >> >> I'm not sure what the purpose would be for an error test in this case, since >> the >> patch is about preventing an error. > > That's what I thought as well, but I think that David was feeling that > there was some need for more extensive testing to make sure that > errors were properly handled instead of segfaulting.
Error messages tend to give an extract about where the interpreter thinks he currently is in its (possibly recursive) input state. This information is not as readily apparent or even extracted by other means. I have no clue about the inner workings of the parser/input. And it appears that those who are deciding whether to commit are in a similar predicament. If the error reactions of the parser appear consistent with expectations, that could provide some basic reassurance that the patch does not work just by chance and does not mess up internal states all too much. -- David Kastrup _______________________________________________ lilypond-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/lilypond-devel
