Le ven. 9 juin 2017 à 16:32, Joerg Schilling < [email protected]> a écrit :
> Yann Régis-Gianas <[email protected]> wrote: > If you plan to use the BNF grammar for the shell that is in the standard, > then it may be of interest that this is most likely not correct. > Actually, we may have found a bug in the BNF grammar. Once corrected, we are using a parsing technology, the menhir parser generator, which allows us to copy the BNF grammar almost verbatim in our development and interpret it in such a way that our parser actually parses all the POSIX shell scripts of our corpus. Being able to include the BNF in our development and document how it should be interpreted (using high-level typed code) is, in our opinion, a strength of our approach, leading us to believe that we are actually implementing the standard as described by the POSIX specification. > A friend did use a program that creates random programs for a given BNF > grammar and none of the created files was accepted by any of the known > shell implementations. > Is this program available somewhere? It would help us complete our testsuite! Thanks for your email. Cheers, -- Yann Régis-Gianas
