| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
|
| |
|
| |
|
|\
| |
| | |
Revert "LaTeX: extend to all projects the #3742 fix for PDF builds at RTD"
|
| |
| |
| |
| | |
This reverts commit 9b2aac68ead15067205470ec28c94af909c080c0.
|
|/
|
|
| |
Fix: #5240
|
|\ |
|
| | |
|
|\ \
| |/ |
|
| | |
|
| | |
|
| | |
|
| | |
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The effect is that indexed Greek letters are sorted with uppercase Greek
letters before lowercase Greek: ΓΔΘ...αβγ...; without this commit the
Greek letters are ordered according to the ascii characters used in the
TeX control sequences as set-up in sphinx.util.texescape, i.e. the order
is αβχΔδεηΓγ... In both cases indexed terms starting with such a Greek
letter end up in the same group as terms starting with a non-letter.
This commit does not create letter-groups for Greek letters, as such
letter-groups would require to make sure non-escaped letters such as α
are accepted by TeX engine, and this requires extra-steps for pdflatex (or
OpenType font with Greek script for xelatex/lualatex), and if those
steps were made default by Sphinx, there would not be a reason anymore
to escape via sphinx.util.texescape the Greek letters from rest source.
|
| | | |
|
| | | |
|
|/ /
| |
| |
| | |
Refs: #5134
|
|\ \
| | |
| | | |
LaTeX: extend to all projects the #3742 fix for PDF builds at RTD
|
| |\ \ |
|
| | | |
| | | |
| | | |
| | | | |
Refs: https://github.com/rtfd/readthedocs.org/issues/2857
|
| |/ /
|/| | |
|
| | | |
|
|\ \ \ |
|
| | | | |
|
| |/ /
| | |
| | |
| | |
| | | |
Also adds usage of \sphinxstyleindexletterhead in python.ist as a
customizable alias of original \bigletter.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
- achieve hyperlinked page numbers in general index, despite xindy's
incompatibility with hyperref,
- obtain compatibility of pdflatex with xindy for documents with
Cyrillic scripts,
- use xindy by default with xelatex/lualatex because any non-ascii
character as first letter of an indexed term breaks PDF build if using
makeindex with lualatex, and produces incoherent index with xelatex.
This commit also switches off xindy logs (which were activated in parent
commits) as they are voluminous and not of immediate use to Sphinx user.
|
| | | |
|
|/ /
| |
| |
| | |
Fix: #5132
|
|\ \
| |/ |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| | |
(I should perhaps have left it standing, as this may cause merge
conflicts if merged after long delay)
|
| |
| |
| |
| | |
closes: #4791
|
|\ \
| |/ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Some packages loaded from user customized 'fontpkg' key may load
textcomp with no option, clashing with sphinx.sty which comes later
and wants it with "warn" option. This makes sure textcomp receives
always "warn" option, even if loaded prior to sphinx.
modified: CHANGES
modified: sphinx/templates/latex/latex.tex_t
modified: sphinx/texinputs/sphinx.sty
|
|/ |
|
| |
|
| |
|
|
|
|
|
|
| |
This is subtle LaTeX thing. Prior to merge of #4370 there was a
`\texttt` which was hiding the potential problem. The fix is to leave a
brace pair in place.
|
|
|
|
|
| |
This make clearer to average LaTeX user the format which is expected for
these macros, in case of redefinitions.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since #2627 (1.4.4), `\code`, and then again at #3116 (1.5)
`\sphinxcode` which is the new name has become more complicated than the
original `\texttt{#1}`. This was to obtain straight quotes in PDF output,
and to allow long inline literals to break across lines.
This means though that users who want to customize `\sphinxcode`, for
example to not only do `\texttt` but to use some colour, have to copy
about 10 lines of complicated LaTeX macros which should be not modified
in any way. This commit moves all the code out of `\sphinxcode` into a
separate macro `\sphinxupquote`.
The LaTeX writer will output `\sphinxcode{\sphinxupquote{foo}}` in place
of former `\sphinxcode{foo}`. Moving the `\texttt` from innermost to
outermost level is with no consequence.
|
|\
| |
| | |
Let LaTeX obey :confval:`math_numfig` for equation numbering
|
| | |
|
|/
|
|
|
|
|
| |
Memo: for Japanese documents, jreport.cls already does that, so this
commit changes nothing. However as the class uses ``\chapter``, this
means that by default howto documents table of contents in PDF have
three levels, whereas manual documents only have two.
|
|
|
|
|
|
|
|
|
|
|
| |
Notes:
- also fixes #4314
- although numbering of figures, tables and code-blocks will be same as
in html, due to issue #4318 the numbering of enclosing sectioning units
themselves may go deeper in html than PDF via latex. But this commit
makes sure numbering goes to minimal depth needed by numfig_secnum_depth
|
|\ |
|
| | |
|