Re: [LyX/master] unicodesymbols fixes.

[Guenter Milde]

Dear Guillaume, dear developers,

On 2016-10-23, Guillaume Munch wrote:
> Le 23/10/2016 à 13:18, Guillaume Munch a écrit :

>> 1942c1942
...
resulted in a change in Math.lyx
...
>> < \textrm{created with \textbf{\textbackslash frac}} &  &
>> \textrm{created with \textbf{\textbackslash cfrac}}\\
>> ---
>>> \textrm{created with \textbf{⧵frac}} &  & \textrm{created with
>> \textbf{⧵cfrac}}\\
...
making it uncompilable.

> As I understand, Günter fixed this at 6ff89c4b9 (I missed it before,
> sorry). I am still curious for the questions below, if anybody knows.

In this case, it was an incomplete change to lib/unicodesymbols (more
details below).  

The intermediate state was not "requiring a fileformat change" but
"buggy".

>> Also, and more importantly, this shows that some changes to
>> lib/[unicode]symbols are in fact file format changes. Is it expected in
>> this case? Is it easy to see which kind of changes to these files are
>> file-format-changing and which ones are not?

>From development.lyx:

  Rule of thumb: 

Whenever there is the danger that a previous version of LyX cannot
open a file using the new feature, a file format update is needed.

  The file format change allows lyx2lyx rules to implement backwards
  compatibility.
  
-> A changed (improved or even buggy) lib/unicodesymbols does not
require a file format change:
  
* lib/unicodesymbols is used on latex export and for relyxing,
  changes will not prevent opening of LyX files.
  
* There is no safe way to backport improvements/changes via lyx2lyx.
  (Converting a literal Unicode character to the correct LICR replacement
  is ugly and has side effects (e.g. prevents the correct XHTML export.)
  
Improvements (additions, simple corrections) just let a literal Unicode
character work that previously did not work properly with 8-bit LaTeX.

For a new lyx version, this is normal and expected: not all improvements
can be backported.

I don't know whether this would be safe for stable, as some documents
that do work with the new revision fail with the old revision.

Günter

Re: [LyX/master] unicodesymbols fixes.

[Guillaume Munch]

Le 23/10/2016 à 13:18, Guillaume Munch a écrit :


1942c1942
< \textrm{created with \textbf{\textbackslash frac}} &  &
\textrm{created with \textbf{\textbackslash cfrac}}\\
---

\textrm{created with \textbf{⧵frac}} &  & \textrm{created with

\textbf{⧵cfrac}}\\



As I understand, Günter fixed this at 6ff89c4b9 (I missed it before,
sorry). I am still curious for the questions below, if anybody knows.



Also, and more importantly, this shows that some changes to
lib/[unicode]symbols are in fact file format changes. Is it expected in
this case? Is it easy to see which kind of changes to these files are
file-format-changing and which ones are not?

Re: [LyX/master] unicodesymbols fixes.

[Guillaume Munch]

Le 08/10/2016 à 16:59, Günter Milde a écrit :

commit efa844702c3844f5bb72ae3218ce175890a67db3
Author: Günter Milde 
Date:   Sat Oct 8 16:57:52 2016 +0200

unicodesymbols fixes.

force=utf8 is required for most characters provided by add-on packgages
and (almost) all mathematical characters, because these are not
set up for inputencs utf8

unicodesymbols.py failed here (python 2.7 under Linux) before the simple fix
included in this commit.




Hi,


This commit results in the following difference in doc/MathMath.lyx
after saving:

1942c1942
< \textrm{created with \textbf{\textbackslash frac}} &  & 
\textrm{created with \textbf{\textbackslash cfrac}}\\

---
> \textrm{created with \textbf{⧵frac}} &  & \textrm{created with 
\textbf{⧵cfrac}}\\

18774c18774
< \mathbf{\int_{n}^{2}f(\theta)=\Gamma}\qquad\textrm{equation with 
\textbackslash mathbf}

---
> \mathbf{\int_{n}^{2}f(\theta)=\Gamma}\qquad\textrm{equation with ⧵mathbf}
18792c18792
< \boldsymbol{\int_{n}^{2}f(\theta)=\Gamma\qquad\textrm{equation with 
\textbackslash boldsymbol}}

---
> \boldsymbol{\int_{n}^{2}f(\theta)=\Gamma\qquad\textrm{equation with 
⧵boldsymbol}}



At first I thought of a bug but if you look closely, \textbackslash is
not replaced by \ but by ⧵ (U+29F5 REVERSE SOLIDUS OPERATOR).

Is this change expected? Is it correct? It seems to result in incorrect
behaviour to me:
* when activating Use non-TeX fonts, it gets output as U+29F5 whereas
U+5C is meant.
* in normal text, \textbackslash remains unchanged by your patch, but
when doing copy-paste from the above formulae to normal text, one gets
U+29F5 (thought it is hidden until one enables Use non-TeX fonts).

Also, and more importantly, this shows that some changes to
lib/[unicode]symbols are in fact file format changes. Is it expected in
this case? Is it easy to see which kind of changes to these files are
file-format-changing and which ones are not?


Thank you,
Guillaume