Regressions

[burrbull]

See below

[burrbull]

1702.00047

[burrbull]

Just open in evince and Ctrl+A, Ctrl+C

On the left is oxidized version.

[crlf0710]

From
https://tt.ente.ninja/#/compare/master-a9958b322b9f46bcbd7e064b156362fe989fd8d1/0ba4b44ac313f8636f4927bd658e07ac3527481d

It seems that 1702.00129 & 1702.00351 are fixed. The first file that's a regression is the 1702.00047 mentioned above. Will investigate into it when i got some time.

[crlf0710]

I compared the pdf outputs of 1702.00047 file, it seems these two type1 fonts:
MSBM10
LMMathExtension10-Regular
are losing one tounicode cmap entry. There's no other differences. Mmmm. Strange.

[Mrmaxmeier]

Here's a minimal reproducer reduced from 1702.00102:

\documentclass{article}
\usepackage{amsmath}
\usepackage{amssymb}

\begin{document}
$\widetilde{\mathbb{P}_{x}}$
\end{document}

[burrbull]

Looking on screenshot above (left oxidized version) it seems to me this is not regression. Looks like something was fixed.

[cormacrelf]

What about it was fixed? All I can look at is the badly kerned “closure” on the left.

[burrbull]

I say about screenshots with widetilde.
What "closure" do you say about?

[crlf0710]

Actually i debugged the code a little the other day. Seemed nothing badly wrong to me. Next time i'll compile a c version and check out how their traces are different.

[burrbull]

@crlf0710 What tools do you use to compare pdfs?

Can you also compare something from https://tt.ente.ninja/#/compare/1fbc8a7e1adedc2a819bffad0af94f3b04c88750/ed980e087cb84b54b9a2e6fc9bd5dab4c2fce275 ?

[crlf0710]

@burrbull To compare pdfs the key point is to force it stop using compressions and binary representations. After that it will be entirely text.
I usually use a tool called mutool from mupdf package to do this.

To convert a file, i run this command.
mutool clean -adfigs <input.pdf> <output.pdf>

After this, i can compare two converted pdfs using any text editor or text diff tool.

[burrbull]

Regression in 1702.04878 fixed in #227

[burrbull]

1702.00102 :

Old~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~New

Here's a minimal reproducer reduced from 1702.00102:

\documentclass{article}
\usepackage{amsmath}
\usepackage{amssymb}

\begin{document}
$\widetilde{\mathbb{P}_{x}}$
\end{document}

Was added between 1002e72 and 5fb2205

Compare: test_agl...crlf0710:test_agl2

[burrbull]

@pkgw

Looks like problem in tectonic was accidentally solved during oxidize.

[burrbull]

@Mrmaxmeier can we somehow filter changed files to see are there other bugs?

Maybe with

grep \widetilde TEX_FILE

or

mutool show PDF_FILE grep | grep /tildewider

Now we have 163 changes:
https://tt.ente.ninja/#/compare/master-a9958b322b9f46bcbd7e064b156362fe989fd8d1/16af24cf690177dc82c4f602ae10e73c70adf65d

[burrbull]

1702.07325
https://tt.ente.ninja/#/compare/master-a9958b322b9f46bcbd7e064b156362fe989fd8d1/9b32a7fb4d3a238d9ddfbdc2feb5f11117082779

Between:
c7e2589...crlf0710:a0fd1d78709922ccbdca911e6689e1b70e7f8324