IRC channel logs


back to list of logs

<daviid>3 warnings and it fails when it enters 'doc', with 0 (%resolve-variable (7 . #{\#t}#) #<directory (guile-useā€¦>) ... Unbound variable: #{\#t}#
<daviid>I actually didn't write any of the guile-lib code, nor the doc, and so far only did 'admin maintaince' - so if any one wants to help, welcome ...
***catonano_ is now known as catonano
<weinholt>daviid, i have some patches for guile-lib, one of them fixes that #{\#t}# problem
<weinholt>daviid, and i'm sure i don't know what goops tutorial you're talking about :)
<daviid>weinholt: i need to rest, but talk tomorow ... tx
***apteryx_ is now known as apteryx
<chrislck>civodul how to remove unbound-variable new between 3.0.2 and 3.0.4 ? compiling gnucash with 3.0.4 causes too many warnings.
<civodul>chrislck: hi! i don't understand the question
<civodul>"new" is unbound?
<chrislck>"new" warnings
<chrislck> shows %auto-compilation-warnings was set precisely to remove unbound warnings for a very long time
<chrislck>since I upgraded ubuntu to 20.10 ie 3.0.2 to 3.0.4 there are now numerous warnings all over the place
<chrislck>IOW up to 3.0.2 no warnings, and now too many warnings
<civodul>oh, what kind of warning?
<civodul>i don't think there are new warning types in 3.0.4
<civodul>however, there's a new "warning level" interface
<civodul>see NEWS
<chrislck>I saw that NEWS, didn't understand the interface
<civodul>settings %auto-compilation-warnings like you do should work, though
<civodul>oh but wait, that's when running "guild compile", right?
<chrislck>yes %auto-compilation-warnings didn't need to be set for each every .scm so far
<civodul>"guild compile" doesn't care about %auto-compilation-warnings
<chrislck>is this a bug then?
<leoprikler>I don't think so. You should be able to pass -Wno-unbound-variables to suppress the warning IIUC.
<leoprikler>or rather passing any set of warnings manually will trigger guild to use your set instead of the default
<chrislck>is there an ENV var we could set to suppress warnings wit guild?
<chrislck>dont want to touch 100s of .scms
<leoprikler>Okay, -Wno does not exist, but -W0 does
<leoprikler>So simply do -W0 in front of your other compilation options and you should be a-ok.
<chrislck>let's try
<chrislck>works \o/
<chrislck>wouldn't be useful to suppress *all* warnings though
<leoprikler>in my personal opinion unbound-variable is one of the most important ones.
<leoprikler>But if you have a module, that does higher tier magic, you may end up using it regardless.
<chrislck>the structure of .scm in gnucash makes it tedious
<chrislck>is it possible to disable unbound-variable alone?
<leoprikler>As I said previously, there sadly is no -Wno
<leoprikler>You'll have to -W0 and then enable all the warnings you want
<chrislck>hm ok
<chrislck>I've filed bug in gnucash: not sure what is the best option
<leoprikler>chrislck: you are missing some obvious options in your listing
<leoprikler>the option to compile gnucash in a manner that makes these variables bound as they should be!
<dsmith-work>Monday Greetings, Guilers
<dsmith-work>daviid: need to add a '?' to query the bot. Hmm. Sadly, it's offline at the moment. :(
<daviid>dsmith-work: ok, i'm pretty sure I didn't use ? when i asked for guile-software, but ...
<daviid>weinholt: ok, I am available, when you are, let me know, we'll fix guile-lib with those patches you have ... tx
<dsmith-work>daviid: You may very well be right. I don't remember right now.
<weinholt>daviid, it's just some small things:
<weinholt>daviid, also interesting perhaps is that i compile md5.scm with -O0 for guile 3.0, because otherwise the tests fail
<daviid>weinholt: ok tx, will look at those - and yes, i did see that the test fails, did you patch the unit-test/, how do you ask to compile -00 for only one file?
<weinholt>daviid, it's a hack in debian/rules (easier to manage than a patch file):
<daviid>weinholt: I wonder how to do that using the autotool chain, i guess it is possible, if anyone knows ...
<weinholt>daviid, i'd try to copy the ".scm.go:" rule as "md5.go:" and add -O0 in the copy. just a guess.
<daviid>weinholt: i'll try that tx
*daviid is abit on and off, sorry, but will get there, i hope :)
<daviid>wingo: hello! on another subject, I need your personal help :), g-golf doesn't compile/run using 3.0.4, because module-use! fails - the module-use! code hasn't changed, but things changed in the module system that make it fail ...
<daviid>wingo: it has been somewhat reported, by dsmith-work, and leoprikler (and dsmith-work ) offered a somewhat 'naive' work-around, but g-golf needs the good old module-use!, could you look at it and tell me if you can and wiil fix the problem? many thanks, the bug is bug#43025
<daviid>wingo: the 'naive' work around is to copy and add the public interface (of a module to another), but 2.2 is a lot more clever :) and holds a pointer to the obarray, so one can dynamicalluy add bindings to the public interface ... g-golf inherently does 'just' that, it reads a typelib, and add bindings to its (g-golf) high level api modules ...
<daviid>wingo: many thanks, i hope you'll find a bit of time, i think this is the only problem g-golf has wrt 3.0.4 - i can't be sure yet, but i think it will work if module-use! works as in 2.2
<daviid>weinholt: do you have asavannah account? would you not prefer to patch guile-lib upstream directly 'yourself' (not being lazy here, just offereing you appear in the upstream git maintaince logs ...
<weinholt>daviid, too busy, writing a USB stack
<daviid>ok, I'll name you of course ... tx
<daviid>weinholt: shall i add you as the copyright author of these patches?
<daviid>in the files themselve i mean
<weinholt>daviid, no, i think the changes are too small for that
<ArneBab>we wingo
<daviid>weinholt: in the patch you wrote about using $(GUILD), you also patch, but that file is generated, isn't it the case on debian as well?
<daviid>*src/ but i don't think we should do that ...
***nckx is now known as ncksneekx
***ncksneekx is now known as nckx