Hugs folk, I thought I'd run the latest Hugs against some of the GHC tests, in particular to test the module system since it's received some attention recently. I've found a few bugs... all these test cases can be found under fptools/testsuite/rename. 1. Hiding constructors doesn't work properly. See rename/prog002, which should fail to compile. 2. Pattern bindings which bind no variables are erroneously disallowed (rename/should_compile/rn003). 3. some problem with qualified imports (rename/should_compile/rn026 and rename/should_compile/rn027). 4. defining a variable which clashes with an import is ok as long as the variable is never referred to. Hugs disallows it. (rename/should_compile/rn029). 5. some problem with empty declaration lists (rename/should_compile/rn034). 6. infix declarations are legal inside a class declaration (rename/should_compile/rn035). 7. precedence parsing with unary negation is wrong. (rename/should_fail/rnfail017.hs). 8. it appears that extra qualified names are being brough into scope (rename/should_fail/rnfail030.hs and rnfail031.hs). 9. hiding doesn't work with qualified imports (rename/should_fail/rnfail032.hs and rnfail033.hs). 10. naming a constructor in an export list is illegal. (rename/should_fail/rnfail035.hs). We really ought to think about merging our test suites at some point; the stuff in fptools/testsuite is supposed to be largely compiler-independednt, but it will need some work to make it go through with Hugs. Cheers, Simon
Thanks, these now behave as expected with Hugs, bar the 2nd one, which is a known Hugs issue. I'm not planning on tackling it right now; others feel free.. --sigbjorn ----- Original Message ----- From: "Simon Marlow" <simonmar@microsoft.com> To: <cvs-hugs@haskell.org> Sent: Wednesday, September 11, 2002 03:07 Subject: Bugs
Hugs folk,
I thought I'd run the latest Hugs against some of the GHC tests, in particular to test the module system since it's received some attention recently. I've found a few bugs... all these test cases can be found under fptools/testsuite/rename.
1. Hiding constructors doesn't work properly. See rename/prog002, which should fail to compile.
2. Pattern bindings which bind no variables are erroneously disallowed (rename/should_compile/rn003).
3. some problem with qualified imports (rename/should_compile/rn026 and rename/should_compile/rn027).
4. defining a variable which clashes with an import is ok as long as the variable is never referred to. Hugs disallows it. (rename/should_compile/rn029).
5. some problem with empty declaration lists (rename/should_compile/rn034).
6. infix declarations are legal inside a class declaration (rename/should_compile/rn035).
7. precedence parsing with unary negation is wrong. (rename/should_fail/rnfail017.hs).
8. it appears that extra qualified names are being brough into scope (rename/should_fail/rnfail030.hs and rnfail031.hs).
9. hiding doesn't work with qualified imports (rename/should_fail/rnfail032.hs and rnfail033.hs).
10. naming a constructor in an export list is illegal. (rename/should_fail/rnfail035.hs).
We really ought to think about merging our test suites at some point; the stuff in fptools/testsuite is supposed to be largely compiler-independednt, but it will need some work to make it go through with Hugs.
Cheers, Simon
Let me correct this slightly - thought I had a good handle on the changes reqd to fix the issue with precedences and unary negation, but alas no. So, no change in the status of item 7. --sigbjorn ----- Original Message ----- From: "Sigbjorn Finne" <sof@galois.com> To: "Simon Marlow" <simonmar@microsoft.com> Cc: <cvs-hugs@haskell.org> Sent: Thursday, September 12, 2002 14:59 Subject: Re: Bugs
Thanks, these now behave as expected with Hugs, bar the 2nd one, which is a known Hugs issue. I'm not planning on tackling it right now; others feel free..
--sigbjorn
----- Original Message ----- From: "Simon Marlow" <simonmar@microsoft.com> To: <cvs-hugs@haskell.org> Sent: Wednesday, September 11, 2002 03:07 Subject: Bugs
Hugs folk,
I thought I'd run the latest Hugs against some of the GHC tests, in particular to test the module system since it's received some attention recently. I've found a few bugs... all these test cases can be found under fptools/testsuite/rename.
1. Hiding constructors doesn't work properly. See rename/prog002, which should fail to compile.
2. Pattern bindings which bind no variables are erroneously disallowed (rename/should_compile/rn003).
3. some problem with qualified imports (rename/should_compile/rn026 and rename/should_compile/rn027).
4. defining a variable which clashes with an import is ok as long as the variable is never referred to. Hugs disallows it. (rename/should_compile/rn029).
5. some problem with empty declaration lists (rename/should_compile/rn034).
6. infix declarations are legal inside a class declaration (rename/should_compile/rn035).
7. precedence parsing with unary negation is wrong. (rename/should_fail/rnfail017.hs).
8. it appears that extra qualified names are being brough into scope (rename/should_fail/rnfail030.hs and rnfail031.hs).
9. hiding doesn't work with qualified imports (rename/should_fail/rnfail032.hs and rnfail033.hs).
10. naming a constructor in an export list is illegal. (rename/should_fail/rnfail035.hs).
We really ought to think about merging our test suites at some point; the stuff in fptools/testsuite is supposed to be largely compiler-independednt, but it will need some work to make it go through with Hugs.
Cheers, Simon
_______________________________________________ Cvs-hugs mailing list Cvs-hugs@haskell.org http://www.haskell.org/mailman/listinfo/cvs-hugs
participants (3)
-
Ross Paterson -
Sigbjorn Finne -
Simon Marlow