***Guest4094 is now known as micro__
<taylanub>This is really weird; when I `load' my source file, it somehow executes an old version of it. I removed my cache, and it actually re-compiled it, and it seems to even have compiled an old version. <taylanub>It's not the file itself, it's a module it loads. <taylanub>And an old version of the module is loaded in the REPL. Heh. <ijp>,reload (that module) <taylanub>Ah thanks, was trying to use `reload-module'. ***efftee is now known as ft
***micro__ is now known as Guest7847
***Gues_____ is now known as Guest3491
***dsmith_ is now known as dsmith
***Guest7847 is now known as micro_
***brendyyn is now known as brendyn
<rlb>dsmith: guile needs 7.2 -- I forgot about that -- debian still only has 7.1. <taylanub>I'll either wait until I'm done with syntactic.scm, or announce it before doing so and ask for recommendations on how to approach that .. <taylanub>Oh and that last bit of documentation .. although the most tricky part, I should definitely finish that. <taylanub>Integrating this with pointers and such, we could get some really nice integration with C APIs, I think. Someone had proposed that at least. Now I need sleep though, sadly it seems like I can only be productive if I stay up until around 6 AM. :P ***cky944 is now known as cky
***Gues_____ is now known as Guest72842
*cky wonders how ijp is doing in Round 1C at the moment. :-) ***sneek_ is now known as sneek
***Gues_____ is now known as Guest44229
<civodul>do you benefit from the French days off? <wingo>i get the equivalent number of days <civodul>i'm asking because May is almost a month off here ;-) <wingo>didn't do the long weekend thing this w/e <wingo>yeah things are a little strange here, my partner's school in geneva has different days; with her schedule june is much more exciting ;) ***Gues_____ is now known as Guest37357
***solarus_ is now known as solarus
<taylanub>Does multiple-values usage require allocation ? <taylanub>Hrm, if so, I think I can swap out some procedures that return multiple values with syntax forms that let-bind given identifiers to the wanted values instead. <taylanub>Ehh .. `values' is a procedure it seems, meaning it must do allocation for its rest-arguments. :\\ <ijp>you're worrying too much <taylanub>Probably .. I just realized the macro solution is not possible. <taylanub>(Because it will be determined at run-time.) <ijp>multiple values are passed on the stack <taylanub>Seems like the only solution is to port the whole thing to macro expansion time, at which point I can allocate as much as I want. I obscured the implementation to this point for no reason then. <taylanub>ijp: But rest-args require allocation, no ? Is there a values2, values3, etc., or so ? <ijp>(compile '(values 1 2 3) #:to 'glil) <ijp>(program () (std-prelude 0 0 #f) (label :LCASE950) (const 1) (const 2) (const 3) (call return/values 3)) <ijp>pushed straight onto the stack <ijp>we can do this for calls to values that are visible in the code <ijp>obviously something like (define (k123 f) (f 1 2 3)) (k123 values) isn't treated like that <ijp>(where k123 is a top-level define) <taylanub>Hrm, I see `values' in ,optimize output; does that mean it will allocate ? <ijp>taylanub: this is done in the translation from tree-il to glil <ijp>which happens after optimisation <ijp>anyway, if you are asking "is $feature slow", you are probably asking the wrong question <taylanub>It happens in the referencing function for bytestructures, and I think it would be great if simple access didn't require allocation .. <wingo>where's the manual again? ;-) <ijp>we autoload it at the repl, but for scripts (system base compile) <taylanub>wingo: Couldn't find it in (info "(guile) Compilation"). :\\ <taylanub>Hrm, that still doesn't find me the module. :P ("(system base ..)" is mentioned in (info "(guile) Compiler Tower") apparently.) <taylanub>Thanks though, still an info noob, didn't know of i. <ijp>you know, there is a procedure index... <taylanub>ijp: The procedure definition doesn't say in what module it is .. <ijp>kids these days, think C-s is some sort of substitute for a good old fashioned index *grumble* *moan* <wingo>if the module isn't mentioned in the right place, please file a bug <wingo>i know this is diverting your original question, but better to answer it once, well, and in the right place <wingo>i spent time on those docs and would prefer to spend the current time hacking on the compiler <wingo>why don't we have exact-integer? in the default environment ***Gues_____ is now known as Guest22624
***ijp` is now known as ijp