On Mon, Feb 24, 2020 at 01:08:13PM -0800, Randy Dunlap wrote: > On 2/24/20 11:12 AM, Matthew Wilcox wrote: > > On Mon, Feb 24, 2020 at 11:58:51AM -0700, Jonathan Corbet wrote: > >> On Mon, 24 Feb 2020 10:52:27 -0800 > >> Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote: > >> > >>> On Mon, Feb 24, 2020 at 07:47:19PM +0100, Jonathan Neuschäfer wrote: > >>>> On Mon, Feb 24, 2020 at 11:08:15AM -0700, Jonathan Corbet wrote: > >>>>> On Sun, 23 Feb 2020 23:22:27 +0100 > >>>>> Jonathan Neuschäfer <j.neuschaefer@xxxxxxx> wrote: > >>>>> > >>>>>> Without double-backticks, Sphinx wrongly turns "--version" into > >>>>>> "–version" with a Unicode EN DASH (U+2013), that is visually easy to > >>>>>> confuse with a single ASCII dash. > >>>>>> > >>>>>> Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@xxxxxxx> > >>>>> > >>>>> This certainly seems worth addressing. But I would *really* rather find > >>>>> a way to tell Sphinx not to do that rather than making all of these > >>>>> tweaks - which we will certainly find ourselves having to do over and > >>>>> over again. I can try to look into that in a bit, but if somebody were > >>>>> to beat me to it ... :) > >>>> > >>>> This seems to do the trick: > >>>> > >>>> diff --git a/Documentation/conf.py b/Documentation/conf.py > >>>> index 3c7bdf4cd31f..8f2a7ae95184 100644 > >>>> --- a/Documentation/conf.py > >>>> +++ b/Documentation/conf.py > >>>> @@ -587,6 +587,9 @@ pdf_documents = [ > >>>> kerneldoc_bin = '../scripts/kernel-doc' > >>>> kerneldoc_srctree = '..' > >>>> > >>>> +# Render -- as two dashes > >>>> +smartquotes = False > >>> > >>> I think what Jon was looking for was the ability to selectively turn > >>> smartquotes off for a section and then reenable it? > >> > >> No that's not what I was thinking, actually. Unless somebody can come up > >> with a good reason to the contrary, just disabling that behavior globally > >> strikes me as the right thing to do. > > > > Well, sometimes -- when the time is right -- I like to use en-dashes. > > It's probably no great loss, though. > > > > grep finds me these interesting examples: > > > > Documentation/RCU/Design/Memory-Ordering/Tree-RCU-Memory-Ordering.rst:Tree RCU's grace--period memory-ordering guarantees rely most heavily on > > Documentation/accounting/psi.rst:scarcity aids users in sizing workloads to hardware--or provisioning > > Documentation/admin-guide/acpi/cppc_sysfs.rst: -r--r--r-- 1 root root 65536 Mar 5 19:38 feedback_ctrs > > Documentation/admin-guide/mm/hugetlbpage.rst:task that modifies ``nr_hugepages``. The default for the allowed nodes--when the > > Documentation/block/null_blk.rst:home_node=[0--nr_nodes]: Default: NUMA_NO_NODE > > Documentation/admin-guide/svga.rst::Copyright: |copy| 1995--1999 Martin Mares, <mj@xxxxxx> > > Documentation/media/uapi/v4l/pixfmt-srggb10-ipu3.rst: G\ :sub:`0108high`\ (bits 1--0) > > > > (in all, 368 lines, but they're not all in .rst files) > > > > Not trying to be contrary, but I would prefer to keep .rst files as much > ASCII as possible. I don't think anybody is arguing otherwise. The question is whether minusminus should be left as a pair of minus signs or whether it should be converted into an en-dash.
