- May 15, 2000
-
-
Simon Marlow authored
wibble
-
Simon Marlow authored
fix bug in module initialization
-
Simon Marlow authored
add bLOCK_SIZE
-
Simon Marlow authored
add CurrentTSO, CurrentNursery, bdescr->start and bdescr->free offsets.
-
Simon Marlow authored
forgot a file in the Upd_frame ==> upd_frame commit.
-
Simon Marlow authored
Change the order of building stubs vs. the main .hc file so that one doesn't trample on the other. This stuff is *so* horrible.
-
Simon Marlow authored
callback is a definite via-C test
-
Reuben Thomas authored
Updated to conform to different output.
-
Reuben Thomas authored
Extra rule to make gmp/gmp.dll.
-
Reuben Thomas authored
Removed commented-out out-of-date code from the test, and updated stderr.
-
Simon Marlow authored
The NCG should now support _ccall_GC (i.e. foreign import safe).
-
Simon Marlow authored
remove extra \n in debug message
-
Reuben Thomas authored
Stdout modified in accordance with output (it's the same as io005.hs, which has itself been modified).
-
Reuben Thomas authored
Reversed some formatting changes to make more similar to last version that worked (currently seems not to work on Windows or Linux).
-
Simon Marlow authored
syslib concurrent depends on lang
-
Simon Marlow authored
back out previous misguided commit.
-
- May 14, 2000
-
-
sven.panne@aedion.de authored
Recurse into docbook subdir, too
-
Jeff Lewis authored
Wobble. Fine tuning tcSimplifyAndCheck a bit further (wrt implicit params). The key is that a method that doesn't constrain a local tyvar, but does has implicit params, needs to be reduced further.
-
- May 13, 2000
-
-
Jeff Lewis authored
A clean-up pass on fundeps and implicit params. Haven't yet incorporated changes from Hugs/GHC meeting yet, tho. - Fixed up several places in Type.lhs where IPNotes were probably being incorrectly handled. Strongly suggests a better solution should be implemented for marking implicit params than piggybacking on NoteTys. - tcSimplifyAndCheck was handling implicit params incorrectly (holding on to them when it should have been booting them out to frees). - Improved improvement WRT type signatures (the signature is now taken into account when improving). - Added improvement when matching against local polymorphic types.
-
- May 12, 2000
-
-
sven.panne@aedion.de authored
Fixed bug in integerTo{Word,Int}64zh primops: They used the number of allocated *words* to determine size of an integer, but this is wrong due to normalization. When e.g. two large numbers with different sign, but approximately same absolute value are added, MPN_NORMALIZE cuts down the *size*, but the allocation stays the same. Simple fix: Use size parameter of macros directly. Easier and more correct! :-)
-
Reuben Thomas authored
Changed jade to $JADE
-
Reuben Thomas authored
Add JadeCmd to find path of jade.
-
Reuben Thomas authored
Corrected TOP.
-
Reuben Thomas authored
Reorganisation to parametrise *.sh files on name of Jade, which is now found by the configure script.
-
Julian Seward authored
DietHEP demo program wurble
-
Julian Seward authored
Cygwin support machinery for manufacturing DietHEP.dll and building a small demo program (dh_demo.exe).
-
Reuben Thomas authored
db2html now parametrised on location of files it needs (GIFs and CSS).
-
Reuben Thomas authored
Added to repository: GIFs and CSS that are copied to HTML directories by db2html.
-
Reuben Thomas authored
New driver.
-
Reuben Thomas authored
Makes db2html from db2html.sh
-
Reuben Thomas authored
db2html is now produced from db2html.sh
-
Julian Seward authored
3rd time lucky with DIET_HEP and Cygwin. I hope.
-
Julian Seward authored
Rename DietHEPpish things HMODULE, GetProcAddress, etc, since those names conflict with Windows builtins.
-
Simon Marlow authored
compile PrelRules with -fvia-C for now.
-
Reuben Thomas authored
Removed the SGML directories stuff. This should all be set up by the user (who won't have to if the SGML tools are correctly installed).
-
Julian Seward authored
Minor wurbles to make it compile on Cygwin following DietHEPpery.
-
Simon Marlow authored
Update for changes in the profile output.
-
Simon Marlow authored
The time/allocation profile now shows costs as both - "individual" (incurred by the specific cost-centre stack), and - "inherited" (incurrent by the cost-centre stack and all its children). In addition, the "inner" column has been removed (should be the same as the sum of the scc counts of the children of the current node), and the "caf" column has been removed (wasn't very useful).
-
Reuben Thomas authored
Clarified comment about SGML_TOP.
-
Reuben Thomas authored
Changed absolute paths to use SGML_TOP.
-