IRC channel logs

2026-06-26.log

back to list of logs

<mwette>I like starting line and column at 1: makes more sense.
<mwette>for reporting
<ekaitz>old: you are looking at the unmatched paren issue, right?
<old>ekaitz: that and more. the reader erro reporting is not great and I want to improve it
<ekaitz>old: i took a look into it when I found the issue and talked with arun. It is designed in a way the error reporting is not really taken into account
<old>it was ported from C to match exactly I think
<old>but we can do better now
<ekaitz>it was trickier than I thought it was going to be
<ekaitz>surely
<old>for example I merge today c19c452f8d5f02b7ff60dfd19055d3c27e0bc398
<old>hmm for the unbalance parenthesis, I did it locally
<old>it's kind of trivial if you keep parameterize context
<old>> c19c452f8d5f02b7ff60dfd19055d3c27e0bc398
<old>#( 1 . 2) => foo.scm:2:3: bad literal vector `#(1 . 2)'. Did you mean `#(1 2)'?
<old>Instead of the old backtrace in the reader
<ekaitz>oh yeah
<ekaitz>that's great
<old>oh I made a mistake. The reader annotate the syntaxes with -1 on column aready okay
<old>so it's already a 0 index base for column cool
<euouae>Hello friends
<euouae>How are things? I have been designing a game in Godot 4 and it's a lot of fun. I think Godot recently got a nice boost in dev due to nonsense with unity. Anyway, how is Hoot doing?
<euouae>I see there's been a new release, 0.9.0
<old>so says we change error location of the reader
<old>to be the begining of the expression instead of the end
<old>would that be considered a breaking change hm ?
<ekaitz>i'd say yes
<old>yet this only change the reported message in error
<old>location of syntax objects and source properties are unchanged
<old>so I guess if someone grep their read-error that could be breaking hm
<old>if that's even a thing
<ekaitz>oh, in that case maybe it isn't?
<rlb>I don't think we've generally included the format of our error messages in our "api", and I suspect we generally don't want to.
<old>right
<old>the error location itself is not expose to the exception
<old>i.e., you can't have the error location as numbers. You need to parse the message to extract them
<old>which I would like to chnage in the future. For good error reporting, origin tracking is critical
<rlb>Yeah, I could easily imagine, all else equal, you might want both the start and the end as attributes/components/whatever of an exception like that.
<dsmith>Hmm. Could Geiser (or other?) be aware of error reporting?
<rlb>How can we tell the scm_t_hash size from scheme? Or alternately, do we have easy access to sizeof(uintptr_t) in scheme?
<rlb>ahh, at least (sizeof uintptr_t) works for (system foreign)
<mwette>(use-modules (system foreign) (sizeof uintptr_t)
<mwette>you beat me to it
<rlb>:)