IRC channel logs


back to list of logs

<canssss>Is there kind of test suite for interpreter of Guile?
<ft>Like the "test-suite" subdirectory in the source tree?
<cky>canssss: What ft said.
***mario-go` is now known as mario-goulart
<civodul>Hello Guilers!
<ArneBab>moin civodul
<wingo>meep meep
<sneek>Welcome back wingo, you have 1 message.
<sneek>wingo, daviid says: that I've sent a request for inclusion for guile-cairo
<civodul>howdy wingo!
<ArneBab>hi wingo
<wleslie>forgot to mention I enjoyed your ghm talk.
<wingo>i realized that i introduced an n^2 thing into the 2.2 compiler
<wingo>which is making it slow down terribly
<wingo>am working on fixing that
<wleslie>oh that's what that tweet was about (:
<wingo>i knew it was n^2 beforehand but i didn't respect that fact ;)
<wleslie>is it an effect analysis thing?
<wingo>everyone has to get burned sometime
<wingo>no, effect analysis is O(n)
<wingo>it's something i haven't written about yet -- (language cps types)
<wleslie>I've done O(n^2) effect analyses before
<wingo>fortunately writing about things is a good forcing function to fix embarassing things elsewhere
<wingo>wleslie: this one is really local -- it just computes effects for small expressions
<wingo>the complexity of e.g. cse is another question
<wingo>i think it's n^2 as well, though the constant factors aren't so bad -- anyway will look at fixing that too
<wleslie>n is pretty small there anyway
<wleslie>I mean, usually
<wleslie>might not be so true with lots of inlining or macros
<wleslie>yet that's the case you stand most to benefit.
<wingo>well if you have 4000 labeled expressions, n^2 is 16e6, which is already not small...
<wingo>functions can get big after inlining and contification and such
<dje42>civodul: parameters in gdb are the way they are. :-( If this is important enough, then I suspect adding support for gdb parameters (not guile parameters) to gdb-guile will be held off indefinitely (barring someone else taking on fixing everywhere that gdb is broken :-)).
<dje42>python parameters support will have to be permanently broken (or at least maintain support for the old broken interface ... we can't change that one)
<civodul>dje42: ah ok, then it's OK to keep the proposed interface
<civodul>i suspected GDB's internal interface wasn't optimal in the first place
<civodul>so yeah, i guess that's fine
<dje42>Hmmm, looking at the python API, it doesn't provide a similar API anyway, so there is no published API to preserve.
<dje42>Still, gdb's internals would have to be fixed - not sure how many parameters depend on the current design.
<wingo>dje42: sorry i haven't been able to do gdb things recently -- been otherwise busy :/
<civodul>dje42: well, yes, but i understand there's a risk of that forcing us to postpone the whole thing
<wingo>i can follow up on weds/thurs if there is something useful for me to do
<wingo>when i get back home
<dje42>I can only get to gdb-guile on weekends anyway ...
<wingo>heh, ok
<wingo>civodul: did you see
<wingo>i like their approach to gc; perhaps we can do something like that eventually
*civodul looks
<wingo>it's a long article with a few parts; gc is just one of them
<wingo>excellent hacking tho :)
<wingo>and the gc thing links to some scheme paper from the 80s ;)
<wingo>i think that approach meshes particularly well with how people use guile's c api
<wingo>and how we work internally
<wingo>anyway, thoughts for another day
<wingo>the gc is a vertiginous black hole :)
<taylanub>it's really funny how on, FF performs best on the Mozilla benchmark, Safari best on the Apple one, and Chrome best on the one by Google.
<civodul>wingo: i think Guile's previous GC was a Bartlett GC :-)
<civodul>now we know the fancy name ;-)
<civodul>the idea looks nice, but it takes a lot of engineering
<civodul>well ok, the previous GC didn't attempt to move objects etc.
<civodul>it wasn't able to tell which objects were referenced from the heap, and which were referenced from the stack
<ArneBab>taylanub: that’s logical: Even if they were honest, they would optimze the browsers for their benchmarks. You have to decide on some metrics after all.
<ArneBab>(I’m not suggesting, that they are honest…)
<wingo>civodul: seems that moving vs non-moving is an important difference ;)
<wingo>and the interior pointer thing is important as well
<civodul>yes, sure
***fangism-ctrl-Z is now known as fangism
<bhattigurjot>I am sorry to bother you all again but I still haven't figured out why my software's "./configure" file is showing
<bhattigurjot>GUILE_PROGS: command not found
<bhattigurjot>GUILE_FLAGS: command not found
<bhattigurjot>I don't understand why this keeps happening?
<bhattigurjot>#1 manually copied the .m4
<bhattigurjot>#2: ran ldconfig to update the caches
<bhattigurjot>these two methods, both failed to provide the solution to me.
<bhattigurjot>it is as if guile.m4 file is not even recognized
<paroneayea>hello, *
<dsmith-work>sneek: later tell bhattigurjot Running ldconfig allows to be found. Has nothing at all to do with the .m4 file.
<sneek>Will do.
<GermanGT>How can I use a macro defined in other scm file, which is not a module. Or I should define a module to do that? Thanks.
<taylanub>GermanGT: optimally you should define a module, but if you really want you can use 'include' or 'load'
<GermanGT>taylanub: I'm trying to update TeXmacs to use guile 2. The first idea come to my mind was define a module. I don't want change to much the code, but I think that use modules is the natural way in scheme. Thanks, I will try defining a module. Although I expect will be more problems.
<taylanub>good luck :)