Ken Williams wrote:
Hi John,
After my last email I reworked the patch so that all the modifications
are in Platform/VMS.pm. Could you confirm that it still works for you
and I haven't broken it?
I will not be able to get to it for at least 10 hours, and it will
probably be a bit longer, but I can answer your questions now.
> Thanks for the patch. A couple questions:
>
> 1) I see both /include and /Include specified in various places in the
> source. Should your s/// be case-insensitive?
Yes it should be case insensitive. Also VMS accepts the shortest unique
abbreviations, so a case insensitive test for "/inc" to end of string or
next qualifier is needed to find all the cases. The full qualifier name
is /INCLUDE_DIRECTORY= , but using that full name runs in to command
line length restrictions.
> 2) Would it be more appropriate to fold the include-dirs from ccflags
> into the user-specified ones, rather than throwing them away?
In the case of the modules being built, both had an include for '[]' in
them, so it did not matter.
For correctness the includes should be merged. Actually several things
should be merged like /DEFINE=(xxx,xxx,xxx).
It gets complicated because the () is only needed if there is more than
one option.
It may be better for VMS to borrow a solution from the GNV package. The
Open Source GNV package on VMS uses a program named GCC.EXE that accepts
the qualifiers in UNIX format and then spawns the VMS C compiler with
them translated. But that type of change may take some time to
integrate in with Perl. It also adds another subprocess, and some VMS
environments may be sensitive to that.
Tonight I will be doing a full build and test run of the current
blead-perl on OpenVMS Alpha 8.2, which will probably not complete until
tomorrow.
I currently have 4 tests including this one to investigate on Blead for
VMS, the other ones are ZLIB and TAR, and the maintainer of ZLIB also
needs someone to test their latest kit.
-John
[EMAIL PROTECTED]
Personal Opinion Only
