Hi,
is anyone using fxp (a XML parser for SML, see [1]) with Poly/ML
(version 4.1.4 from sf.net)?
I "ported" fxp more or less successfully to Poly/ML 4.1.4 (i.e., it
compiles without errors) but I'm getting strange runtime errors. The
same code runs flawlessly using sml/NJ and MLton, thus I think the
reason of my problem is a behavioural difference of Poly/ML in the
basis library. Finding this difference seems to be not that easy, thus
I'm interested in people that are already using fxp together with
Poly/ML. Do you have ny hints, tipps, or tricks?
Achim
[1] http://atseidl2.informatik.tu-muenchen.de/~berlea/Fxp/
David.Matthews(a)prolingua.co.uk wrote:
> I have built a new release for Poly/ML and installed it on the
> sourceforge site at http://sourceforge.net/projects/polyml . There is a
> database for each of the i386, PPC and Sparc architectures and sources
> for the driver, compiler and Basis library. As before, you need the
> driver plus the appropriate database.
I am about to try this with Isabelle2005. Concerning Solaris: How do you
actually compile the poly driver? Using gcc 3.2.2 produces the following
error:
gcc -DGCC -DSPARC -DSOLARIS -DSOLARIS2 -DBIGHEAP -c -o sparc_assembly.o sparc_assembly.S
sparc_assembly.S: Assembler messages:
sparc_assembly.S:1782: Error: Illegal operands
make[1]: *** [sparc_assembly.o] Error 1
make[1]: Leaving directory
`/export/T3/proj/polyml/polyml-4.1.4-test/driver'
make: *** [none] Error 2
Note that the same problem occurs with the 4.1.3 driver.
Makarius
I have built a new release for Poly/ML and installed it on the
sourceforge site at http://sourceforge.net/projects/polyml . There is a
database for each of the i386, PPC and Sparc architectures and sources
for the driver, compiler and Basis library. As before, you need the
driver plus the appropriate database.
This version has fixes for various bugs including the Linux segmentation
fault which was fixed in the development sources. It also turns off
optimisation in the C compiler which seems to cause problems. The main
change, though, is that the Basis Library now conforms to Gansner and
Reppy's book. This contains a few changes which are incompatible with
the old version, notably the use of array and vector slices.
Please try out this version and let me know if there are any significant
problems. If I don't hear anything in the next week or so I will update
the web site to point to this as the current official release. If I've
failed to include a fix/suggestion you sent me I apologise. I'll try
and make sure these get into the CVS.
I have not built any packages at the moment. If anyone has experience
of building packages for particular operating systems and distributions
it might be better if they could do it. If you do and want to include
it on the sourceforge site let me know. Now that Poly/ML is licensed
under LGPL it should be possible to include it in the operating system
distributions. I will make a Windows package and upload that shortly.
I know it's been a long time since the last Poly/ML distribution but I
think on the whole it's better not to produce them unless there is a
significant reason.
David.