[வெள்ளி ஏப்ரல் 22, 2022] Eli Zaretskii wrote: Hello Eli, Robert >> From: Visuwesh >> Cc: rpluim@gmail.com, 54646@debbugs.gnu.org >> Date: Fri, 22 Apr 2022 16:16:07 +0530 >> >> >> and the GOOD case for bb9 bbf is >> >> >> >> lbearing 1 rbearing 28 width 28 ascent 9 descent 5 >> >> 1 bb9 bbf xadvance 1809 yadvance 0 xoffset 0 yoffset 0 >> >> xpos 0 yoff 0 wadjust 28 >> >> >> >> and the BAD case is >> >> >> >> lbearing 1 rbearing 28 width 28 ascent 9 descent 5 >> >> 1 bb9 bbf xadvance 3193 yadvance 0 xoffset 0 yoffset 0 >> >> xpos 0 yoff 0 wadjust 50 >> > >> > This looks like HarfBuzz is feeding us incorrect data for some reason, >> > but I cannot imagine what that reason could be, sorry. >> >> Could it be some kind of cache? Because when Emacs rendered the entire >> webpage, I was unable to reproduce the issue. > > When Emacs have rendered the entire page, we no longer call HarfBuzz, > because the composition data is indeed cached by Emacs. > > The print-outs you produced are from calling HarfBuzz, so our caching > cannot affect that, AFAIU. I think this might be a cairo+pango problem. My suspicion is due to this bug report https://github.com/harfbuzz/harfbuzz/issues/1892 -- although I don't see the problem with English text as shown in the screenshots. I tried to turn off font metrics in cairo by applying the following patch,