RE: GHC 5.01 "Zarjaz": Test release for Windows

Here's my first feedback concerning GHC 5.01 for Win: There are significant improvements of compilation speed (about 20% faster, sometimes even more) and the generated code's size (about 40% smaller). And the new "--make" option and GHCi are really great features! However, I already discovered some smaller problems, too: - The included version of HaXml (including dtdtohaskell.exe etc.) is not up-to-date. - I had some "Dr. Watson"s when leaving GHCi. I think, it was in a case with a possible heap overflow. I'll try to analyze it further. - When running GHC with the "--make" option, sometimes it stoppes with an error message ("the impossible happened" ...). I will also try to make this situations reproducable. Some smaller problems are: - When leaving out the "-o" option, GHC4.08.2 defaulted to "main.exe", but GHC5 doesn't anymore (produces "a.out"). - I didn't find a way yet to force GHCi to load a module in "interpreted" mode, even if an up-to-date object file is present. Is there a way to do this without deleting .hi/.o first? Christian

- The included version of HaXml (including dtdtohaskell.exe etc.) is not up-to-date.
That's up to HaXml's author, Malcolm Wallace. The InstallShield shipped with the latest version from CVS. (Malcolm, could you possibly update, please?)
- When leaving out the "-o" option, GHC4.08.2 defaulted to "main.exe", but GHC5 doesn't anymore (produces "a.out").
I'll fix that.
- I didn't find a way yet to force GHCi to load a module in "interpreted" mode, even if an up-to-date object file is present. Is there a way to do this without deleting .hi/.o first?
You could touch the .hs file. -- http://sc3d.org/rrt/ | humour, n. unexpected recognition
participants (2)
-
Lescher Christian
-
Reuben Thomas