--- Leopold Toetsch via RT <[EMAIL PROTECTED]> wrote: > Clement Cherlin wrote: > > --- Leopold Toetsch via RT <[EMAIL PROTECTED]> > wrote: > > >>Doesn't a) perl take forward slashes too and b) harness do the glob > > >>anyway? If the shell globs the test args, we'll run into > commandline > >>length issues sooner or later. > >> > >>leo > > > > > > Perl will accept both forward and backward slashes on Win32, and > the > > harness does do the globbing. My patch only deals with the fact > that > > make is treating "\*" as an escaped asterisk, and unescaping it to > just > > "*". I add an additional backslash, thus escaping the backslash > instead > > of the asterisk. Make translates "\\*" to "\*" and all is well. > > Yes. But shouldn't in that case the patch convert the backslashes to > just forward slahes - or much better not backslash it in the first > place > (and possibly quote the *.t args)? > > The patch reverses a wrong action (backslashing) that happens earlier > > somewhere in config. > > leo
The Makefile uses backslashes because backslashes are the native path seperator on Windows, and the command.com shell will not accept forward slashes. My earlier patch intentionally enabled the use of backslashes in Makefiles when compiling with MinGW, in order to get Parrot to compile at all. Parrot now compiles, but this problem with "make test" resulted, and I provided a workaround. I agree that it's a somewhat ugly workaround, but it was the simplest method I could think of. Can you suggest an alternate solution?