We had a meeting of the Mercury group on Fri July 20, 2001. Attendees: Zoltan Somogyi (zs) Simon Taylor (stayl) David Overton (dmo) Mark Brown (dougl) Peter Stuckey (pjs) David Jeffery (dgj) 1. CVS issues - It would be better for CVS log messages relating to branch merges to be placed in a separate log, since they take up a lot of space but give little information. - Files tend to get put into the papers directory with permissions set incorrectly. We should make the permissions check part of the CVS commit script also apply to the papers directory. 2. Documenting introduced names Any new and existing procedure names that are introduced by the compiler need to be documented. The suggested place is a single location in the compiler/notes directory. 3. Ignore counts for breakpoints Zoltan suggested an improvement to the mdb breakpoint mechanism to allow a breakpoint to be ignored for a certain count of times before it is triggered. One issue raised by this is that it would be easy for a user to forget that each call to a det/semidet procedure generates two events, so if they want the Nth call then the ignore count would need to be 2N. Of course, for multi/nondet code the situation is far worse. A solution proposed by Zoltan is to allow options which specify, for the purposes of the ignore count, which events should be counted. For example, the break command could accept the following options: -I N ignore N interface events -E N ignore N entries into procedures 4. Progress reports We went around the table, with each person in turn discussing what they had been working on recently and/or what they were planning to work on. Zoltan Somogyi (zs): - Fixed some bugs in the deep profiler relating to the purity of inserted code. All inserted code must be labelled impure and impurity must be propagated properly. - Has discussed with Fergus a change to stack slot allocation, and has been working on implementing it. - Did some work on the deep profiler paper. - Has done some work implementing the aforementioned ignore counts for breakpoints. Simon Taylor (stayl): - Worked on mmake in the compiler. David Overton (dmo): - Started work on a new paper on mode constraints. - Tried out the new deep profiler, for example by testing it on the compiler itself. Mark Brown (dougl): - Investigated the .net profiling API, which looks more useful for implementing the retry log than the debug API. Although the retry mechanism is intended to be operated from within a debugger, the way the retry log should be constructed has more in common with how a profile is constructed than how a debugger operates. - Investigated the source code for portable.net. David Jeffery (dgj): - Fixing some bugs with removing unused typeclass info arguments from calls into C code. - Fixed a bug in deconstruct. Minutes taken by Mark Brown (dougl@cs.mu.oz.au)