On Sat, 29 Nov 2025 15:59:42 +0100 =?UTF-8?Q?S=C3=A9bastien_Noel?=
<[email protected]> wrote:
On 11/28/25 20:50, Gioele Barabucci wrote:
> But, if I had to guess, I'd say that a changed timestamp somewhere is
> responsible for the discrepancies.
Yes, you are right, timestamps are involved. But it seems they're not
the only culprits. I'm attaching the output of diffoscope.
Hi,
I see three unreproducibility causes:
1. The timestamps, probably the easiest thing to fix. I'll have a look
at them once htic clears NEW.
2. Various checksums being different. That's to be espected, given that
the file contents are different. This will fix itself as soon as the
other issues are resolved.
3. Many things are off by a few bytes, so all their addresses are wrong.
Same as 2, caused by 4.
4. There are tiny differences in the generated font outlines. For
example, in the TTF file
```
<TTGlyph name="u1F61B" xMin="75" yMin="-30" xMax="835" yMax="690">
<contour>
<pt x="75" y="90" on="1"/>
- <pt x="75" y="90" on="0"/>
<pt x="75" y="570" on="1"/>
```
Issues of type 4 are hard to debug and time consuming to fix. Maybe the
"official" font files are generated by another version (a bug has been
added/fixed in the past 5 years?). Or maybe a slightly different version
of the source files, perhaps without visible differences, have been used
to create the official font files. This will require some further study.
However only issue 1 has to be fixed before releasing Debian version -1
of the font (to achieve self-reproducibility). The rest (reproducibility
of the original file and RFN licensing) can be fixed later.
Regards,
--
Gioele Barabucci
