|
From: | Jean Abou Samra |
Subject: | Re: Guile 2 (was: Cairo plans) |
Date: | Mon, 6 Sep 2021 00:07:16 +0200 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.11.0 |
Hm, when Harm brought up this point in the previous thread in May, I did not understand this to be a critical showstopper for adoption of Guile 2.2 (is it?) or I would have prioritized this over working on the new infrastructure to compile binaries.
Not critical: it's not like Guile 1 is good at error locations either or errors in general. But one does encounter a number of cases where using Guile 1 in LilyPond yields better errors -- at least this has been my main source of frustration since starting to build with Guile 2. I guess everybody has their particular point they would want to see improved before switching to Guile 2.
IIRC Harm's comments were mostly about not confusing the average user with ugly backtraces of Scheme internals - which was solved some time ago by only enabling them in debug mode. For the error locations, I had some WIP solution for ~80% of the cases (typos and wrong argument types) by installing a custom exception handler and manually adding the location information of where the parsed code came from. I had hoped to revisit this approach over the weekend, but it didn't happen. I cannot foresee when I will have time for this, but just to let people know that I don't think all hope is lost here.
Nice to hear! Best, Jean
[Prev in Thread] | Current Thread | [Next in Thread] |