[Monotone-devel] sunpro status report

[Patrick Mauritz]

hello all,

based on njs' list, here's my final status report for now (with some
additional issues covered):

removed files in m4/ and elsewhere: they're gone for good. if
interested, someone (just maybe me) could look if there are even more
such files to cover, I only removed those that seemed to differ in my
version of autotools and thus appeared in monotone status.

-Wall is back - I adapted the m4 macro Peter proposed to work with c++
for gcc and studio (which has warnings by default, and only "useless"
warnings in the extra verbose modes), all other compilers in that macro
will most likely take some work to do something reasonable, but I kept
them as help, and it won't hurt.

MM(): I check for typeof() now, and use the boost alternative to
__PRETTY_FUNCTION__, so we're as flexible as possible as long as we rely
on typeof(). patches for an alternative information path (the
no-typeof() path is quiet) welcome!

explicit new[]/delete[]. in one case, I could safely make the array size
constant again, in the other two, there's now a smart pointer for array
from boost around it. slightly nicer code.

botan changes: reviewed and changed appropriately.

revision_set::revision_set(revision_set const &) works better now

[e]?grep -q usage in integration tests - it's now replaced by two m4
macros, QGREP and QEGREP and more portable that way. there are some more
issues, but I'll do them later.

unittests: don't work for now due to a bug in sunpro that affects
boost::random (used by xdelta.cc's unit tests). the sunpro people are
working on that.


I think, I got rid of all the gccisms by now, so it might be worth
trying with msvc again.


patrick mauritz