Luca's meaningless thoughts

I'm on Windows, not sure if this applies, but have you recompiled the stardard library with symbols and stack frames? Disabling inlining also improves stack traces' quality. Also, there's ddbg (again, Windows-only) which worked for my GC-debugging needs.

Tango trunk is supposed to compile again :)

"and you can't get a decent backtrace ever using a debugger "

Have you tried a backtrace hack?

http://petermodzelewski.blogspot.com/2008/01/development-in-d-with-use-of-backtrace.html

http://petermodzelewski.blogspot.com/2008/01/installing-tango-backtrace-hack-on.html

http://petermodzelewski.blogspot.com/search/label/backtrace hack

unit tests are useless, they depend on your code to run

please elaborate on that. Maybe __traits in D2 could be somewhat useful.

Hmm.. we definitly need some more tutorials on unit tests in D.. Is anyone satisfied here with built-in compiler's unittest ? I'm somewhat jealous on what Eclipse support has for JUnit and alike. Descent really needs to have support DUnit http://dsource.org/projects/dmocks/wiki/DUnit ported :))

What's current common knowledge about unittest's support in D? I heard some use flectioned + experimental Descent branch, some use their own testing framework (i believe team0xf has something in-house), or just anyone is happy with -unittest and console output?

Add that to the fact that GNU Gold doesn't work, DMD doesn't work, Tango doesn't work [*] and LDC doesn't work, and that it's already hard to debug in D because most of the mainstream tools (gdb, binutils, valgrind) don't support the language (can't demangle D symbols for instance) and you end up with a very hostile environment to work with.

just wait some time while svn trunk will stabilize things. And for "demangle D symbols", there are gdb patches in Gentoo D-overlay http://www.assembla.com/wiki/show/d-overlay which could be useful. Patches are from zero bugs debugger, i guess. gdc (from svn) seems works fine, maybe other projects will backport them, too.

Well, I guess all you have to do to get some comments in a blog is rant a little =)

by Vladimir Panteleev

I'm on Windows, not sure if this applies, but have you recompiled the stardard library with symbols and stack frames? Disabling inlining also improves stack traces' quality. Also, there's ddbg (again, Windows-only) which worked for my GC-debugging needs.

I've tried compiling with debug symbols and no optimizations and I have no luck still. And I'm on Linux, so I can't use the Windows stuff, but thanks for the recommendations =)

by Lars Ivar Igesund on 2009-05-04 15:55

Tango trunk is supposed to compile again :)

Yei! That's nice, thanks Lars...

by tweak on 2009-05-06 12:05

Have you tried a backtrace hack?

Nope.

by tweak on 2009-05-06 12:05

unit tests are useless, they depend on your code to run

please elaborate on that. Maybe __traits in D2 could be somewhat useful.

I mean, I can't use D unittest'ing because the runtime uses the GC to allocate memory to run the unit tests themselves, so when there is trouble, there's nothing a unit test can do to help me figure what's wrong.

by tweak on 2009-05-06 12:46

And for "demangle D symbols", there are gdb patches in Gentoo D-overlay http://www.assembla.com/wiki/show/d-overlay which could be useful. Patches are from zero bugs debugger, i guess. gdc (from svn) seems works fine, maybe other projects will backport them, too.

I knew there were some patches somewhere, but I would be nice if the patches would be merged upstream, is really a PITA having to bootstrap almost everything patched to work with D. It's a butchery job =)

I'd love to have time to make pressure to gdb to accept D support patches, or make patches to demangle D symbols in Valgrind and nm, but unfortunately I have to finish my thesis first :S (which by the way, i would be much more easy if that get done =P)

Druntime builds both release and debug libraries, and you can link against the proper one with a bit of work. Assuming you're using D 2.0, that is.

I'm using D 1.0 (LDC), but I know Tango runtime can do that too.

I'd love to have time to make pressure to gdb to accept D support patches, or make patches to demangle D symbols in Valgrind and nm, but unfortunately I have to finish my thesis first :S (which by the way, i would be much more easy if that get done =P)

well, there are alot of demangle.d versions floating around, just google it (one is in dmd*.zip). So it would be possible to check gdb patches and repeat the same for nm and valgrind.

Have you tried a backtrace hack?

Nope.

I guess backtrace hack is for dmd-windows only, so it would take some effort to port it to Linux, anyway. Also, xfLinker from team 0xf could be somewhat useful to port that hack, but that hack seems to be windows-specific at first glance.

I mean, I can't use D unittest'ing because the runtime uses the GC to allocate memory to run the unit tests themselves, so when there is trouble, there's nothing a unit test can do to help me figure what's wrong.

so, static assert + your own new/delete from global static array is your best friend?