From: Craig Chambers (chambers@cs.washington.edu)
Date: Fri Dec 13 2002 - 15:38:26 PST
I'm publishing some more things.
*) Most importantly, I'm fixing a subtle bug with my previous publish
to make results visible to the debugger. Several tricky issues
cropped up when I compiled the compiler with the compiler; the
resulting compiler crashed in an optimization routine. The code is
now much more "normal" with how it treats result values; the old
Vortex compiled had some hacks that turned out to be fragile when I
made my (nominally unrelated) changes to make results visible.
*) I fixed a small glitch to my next and finish breakpoint code. (I
used these facilities when debugging the above problem!)
*) I have noticed that recently, I can't run Vortex under gdb, because
Vortex's unaligned signals (supporting fast access to tagged data)
were not being handled, and so were crashing Vortex. I got
sufficiently irritated that I investigated, and implemented a patch.
(I think the problem is in gdb; it's sending the sigsegv signal to
Vortex in place of sigbus.)
*) There are a couple of convenient functions in the run-time system
for printing Cecil objects. You can invoke pp(addr) on any Cecil OOP
or Cecil run-time data structure, and pp will print it out in
low-level format. I wanted a way to invoke the Cecil object's
print_string method from gdb, so I added it: pr(obj) on a Cecil OOP
will send it print_string, and then send that result print_line.
Very nice.
*) I made some small tweaks to scripts related to building releases.
I've also been working on several things related to a possible
upcoming release, including sprucing up manuals and their make scripts
and starting a section on the internal web page documenting the steps
in doing a release.
I built a new vortex compiler from the compiler, built whirlwind from
that, built a simple whirlwind test program from that, and also built
a no-debug version of vortex and ran make-profile using it. Not
exhaustive testing, but reasonable. (I'm a little nervous about
releasing the system with only lightly tested executables....)
-- Craig
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://mailman.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2.1.5 : Fri Dec 13 2002 - 15:38:31 PST