Em Thu Feb 4, 2021 at 8:28 PM -03, Jonathan Corbet escreveu: > > Nícolas F. R. A. Prado <nfraprado@xxxxxxxxxxxxxx> writes: > > > As Jon previously noted [1], it would be nice if automarkup supported relative > > paths as well when cross-referencing to other documents. This adds the support > > for it, and documents it. > > > > Jon, after applying this, 43bc3ed73639 ("docs: dt: Use full path to enable > > cross-reference") could be reverted without the link stopping to work. > > > > [1] https://lore.kernel.org/linux-doc/20201130142404.6ef6689e@xxxxxxx/ > > > > Nícolas F. R. A. Prado (2): > > docs: Enable usage of relative paths to docs on automarkup > > docs: Document cross-referencing using relative path > > > > Documentation/doc-guide/sphinx.rst | 30 ++++++++++++++++++++---------- > > Documentation/sphinx/automarkup.py | 7 +++++-- > > 2 files changed, 25 insertions(+), 12 deletions(-) > > So I finally got around to playing with this set. One thing I found is > that some of the references that were being caught before were not > now... after far too much time, I figured out that the problem was > references to .txt files, of which we have quite a few in the docs. > admin-guide/kernel-parameters.txt in particular is quite popular. > Before this change, those were being turned into xrefs, afterward not. Hm, but what is actually being linked to is the .rst of same name, admin-guide/kernel-parameters.rst. Both when adding a file to the Sphinx's index, as well as when cross-referencing, the file name is used without the extension, because the .rst extension is implied. In the case of admin-guide/kernel-parameters.txt though, the .rst file includes the .txt file, so it actually makes sense to cross-reference to the .rst when writing the path to the .txt. But in the case of core-api/refcount-vs-atomic.rst, for example, where there is a reference to memory-barriers.txt, which is a file with no .rst counterpart, it will fail to cross-reference. At the moment that's harmless, but it will become a problem after we enable warnings. > > To address that, I applied this little tweak: > > -RE_doc = re.compile(r'(\bDocumentation/)?((\.\./)*[\w\-/]+)\.rst') > +RE_doc = re.compile(r'(\bDocumentation/)?((\.\./)*[\w\-/]+)\.(rst|txt)') > > That seems to make things work properly again. > > While tracking this down I put in a print for failing cross references, > and noted that we have quite a few; it's a useful way to see where the > stale references are. Maybe I'll try to hack together something to make > those stand out so we can fix them. That was already on my backlog, so I could probably do it in the next few days :) (if you don't get to it first, of course). Thanks, Nícolas