Older blog entries for pvaneynd (starting at number 17)

Still having problems with lazy allocation mode. Sigh. Why it suddenly decided to bit-rot on me is beyond me...

So now I'm trying to understand why it segfaults at eip=0x0 by pure meditation

Sigh. Just as was buzy porting the new CLX to sbcl (actually porting some socket interface of sbcl to clocc-port) I find out that cmucl's lazy mode fails on the new 2.4.7 kernel.


So now I'm again debugging OS/GC interactions... Fun fun fun

fufie: logical pathnames are a bitch. I see that in a few years we will do away with them in cCLan, but for the moment let's try to make use of them so we again a little experience of what it actually needed...

The fast that the clisp people have their own special view on the standard doesn't help of course :-)

What have I been doing? Mostly trying to get sbcl working with the new CLX I've been working on, and discovering protocol errors in said POS.

Arrived at LSM at last... First impressions: the south of france looks a lot more like Italy then the north of france. The change really is impressive

At the moment I'm just keeping up with debian in the DHCP 'room'... And when I see how much the others are working I might have trouble with keeping up :-). Now to find the people I know from email in 'real' life... and to prepare that talk...

Found it! After searching high and low I finally found the error that make cmucl crash: bad FFI-GC voodoo.

Now just fix it, test and release a bunch of new packages!

Sigh. Just as I tought I had fixed the problems they reappear. Just this time it's more difficult: the build only works if I'm strace-ing the lisp! So when I try to find out why it crashes it doesn't!


This just after I found out that the hemlock and clue packages have slight problems....I'll try to put the stuff that works into CVS tonight (CET here), so that the others can play with it a bit. Those problems and the fact that I foolishly agreed to give a presentation at LSM makes my life a bit very buzy. Oh and of course now I'm reading another Ian M. Banks book, just to round thing off :-)

Aarrgh I found why it suddenly broke: I used

cvs -q update -APd .

and that introduced an incompatible change to the layout of the logical hostname structures!

So at least I know why it goes wrong now :-) So now I can just check if my corrections worked and then time-travel forward until I hit problems...

Pfff. How many times I've been stabbed in the back with those incompatible changes... I long to have cmucl stable and to start working on sbcl a bit more...

Still no luck in the bugsearch, but I just wanted people to know that because master.debian.org is down I can't read my email, so send time-critital things to my yahoo.com account (same username as on debian).

Just as I'm ready to release a new version of common-lisp-controller I'm stuck with the lazy memory-allocation code. Bummer.

The fact that I cannot use gdb, that it happens only now and again and that it seems related to the gc only makes the task more interesting.

At least there is one piece of good news: my brother visited 'Little Italy' in NY and got me a T-shirt with the text: Pray for me, my wife it Italian

Long time no diary entry... I've been buzy with common-lisp-controller, I've made it a bit more feature-complete and robust. It is amazing how many errors there can be in a few lines of shell-scripts :-(. And now that I'm close to releasing it cmucl bombs at random when I use lazy memory allocation. Bummer. Now that I've found out that not only doesn't cmucl handle large memory machines its default configuration, it also is allergic to SMP machines. Using lazy seems the only sollution, so I need to find this damn bug! Life: sucks a lot less then a year ago, but I would like a little more time to read... I've got a stack of books a meter high I need to read. Sigh. Liking C'T and iX doens't help: they amount to almost 1000 pages of interesting stuff a month. Holidays: I'm going to debian-one it seems, and we're going to drop by a friend who is now doing a post-doc in Boston. So much to see/do so little time...

8 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!