IRC channel logs


back to list of logs

<rekado>example of refresh recommendations:
<rekado>all the recommendations about r-dt are wrong, of course, because these are JavaScript inputs
<rekado>they do need to be checked, though
<rekado>sometimes there will be a recommendation to move an input from propagated-inputs to inputs; we also need a way to override this.
<rekado>for cases like the r-dt upgrade I usually remove the snippet and get the sources for both versions (the current and the updated one); then diff the JS sources briefly
<civodul>rekado: i started updated the patch set yesterday to include properties, as discussed the other day
<civodul>so there's updater-extra-inputs and updater-ignored-inputs
<civodul>and likewise for propagated and native
<civodul>i've added them to all the packages that had a comment about r-knitr
<civodul>for r-dt, we get this:
<civodul>all the JS is gone :-)
<civodul>interestingly, if we rewrite in "gexp style", these things would probably move from inputs to arguments
<rekado>how come the propagated inputs are no longer sorted?
<civodul>oh, good point!
<civodul>lemme see
<rekado>what does moving from inputs to arguments mean in this case?
<civodul>instead of referring to inputs by label, you'd directly refer to them in the gexp
<civodul>as in: (for-each ... '#$javascript-sources)
<rekado>I found this difficult in some cases because we don’t have a way to find inputs by suffix
<rekado>e.g. when the input is just a single JavaScript file then it will have a name like /gnu/store/…-the-file.js
<rekado>but I cannot know the prefix
<civodul>how about (strip-store-file-name "/gnu/store/…-the-file.js") => "the-file.js" ?
<civodul>rekado: v2 here:
<civodul>hasn't fully arrived yet