Karl Williamson wrote: > The first thing that seems at all suspicious to me is at line #8458, > where the os390 'value' becomes negative. I have no idea what 'value' > means, but I suspect it should not be negative.
Each token emitted by toke.c can have a value associated with it. Only some tokens use the value. In those cases the value in the struct that yylex returns may or may not have been written to. Most likely it is a memory address left over from a previous token. (The same struct is reused over and over.) This value is probably harm- less. Is the value something arbitrarily large on Linux? > Where things really start to go wrong is at line #8696. The linux is > looking at "my ..." and os390 is looking at "\nmy ..." I'm doing this blindly. Is this the 'my' in the here-doc? $WANT = <<'EOT'; my $VAR1 = '42 '; EOT The here-doc parser is rather complex, so bugs lurking in there would not surprise me.
