Re: pr19.test

[Tom Tromey]

>>>>> "edward" == edward  <address@hidden> writes:

edward> On systems with make programs which delete intermediate
edward> targets (including cygwin), I expect pr19 to fail with foo.c
edward> not found.

I thought we circumvented this by emitting dependencies so that
intermediate files weren't deleted.  Do we not?  Maybe I'm really
confused... it's happened before.

edward> On cygwin itself, there is an additional problem due to the
edward> "make clean" test failing. foo.exe isn't removed
edward> properly. That can be fixed by putting AC_EXEEXT into
edward> configure.in

Yes, this is a real problem.  The test suite should probably be
required to use AC_EXEEXT.  Bummer.  Could you submit a PR about this?

edward> i suppose lang_*_rewrite should at least check for
edward> noinst/check targets before pushing into dist common, similar
edward> to what happens in am_install_var. that's a separate issue
edward> though =)

I don't agree.  I think where something is installed is orthogonal to
whether it is built.  It makes perfect sense to distribute noinst_ or
check_ sources.  It also makes sense not to, sometimes.  That is why
we introduced the dist_ and nodist_ prefixes.

Tom