Discarded glues at the top of a frame shall not account in glue adjustement #1752
Conversation
Looking closely, this test expection had an extra parskip at the bottom of the first page, exactly the same symptom as sile-typesetter#1647.
It's harder to say how it relates to sile-typesetter#1647, but checking with -d all, this test showed a small vertical space at the bottom of page 3, meaning the content frame was not stretched as much as expected. It seems to be some lineskip. Now the text seems fully stretched within the content frame.
The changes from sile-typesetter#1647 impact this test on page 3 (last line is slightly lower), this doesn't seem worse though the test is hard to check.
|
I tried this on one of my book project where I was having some troubles and it seems to be a solid improvement. I don't like the way settings and variables are scoped, but at least this code makes it explicit what is being tracked where and what subsystems need to interface with others where. I think this will make a settings rescope involving typesetters and frames easier in the future. Is there any reason we can't bump this up to v0.14.9? I'll look into the |
|
301 has always been a hard bug to replicate. It looks like the change in it is for the better in line with the other affected tests you described. It is still showing something that looks like the same bug, but it actually is different. The unfilled frames are the way they are because of the non-stretchable skips with specific metrics used to trigger the bug. |
Nope, just added it. It was 2AM or so when I completed it, I must have been too tired to notice I had the milestone on the issue but not on the related PR ^^ |
Closes #1647
It turned out to be a bit trickier than I thought and needs scrutiny...
Regarding impacted tests (besides a dedicated test for that case)
bibtex.sil (using parskip) was actually revealing the same issue.
The extra space is no longer showing and the page is maximally stretched.
bug-252 is impacted on page 3, the small space below now disappears. Seems better.
bug-301 is much harder to judge. It doesn't seem worse, but it's hard to tell.