scripts/get_abi.py: add support for undefined ABIs

The undefined logic is complex and has lots of magic on it.

Implement it, using the same algorithm we have at get_abi.pl. Yet,
some tweaks to opti

scripts/get_abi.py: add support for undefined ABIs

The undefined logic is complex and has lots of magic on it.

Implement it, using the same algorithm we have at get_abi.pl. Yet,
some tweaks to optimize performance and to make the code simpler
were added here:
- at the perl version, the tree graph had loops, so we had to
use BFS to traverse it. On this version, the graph is a tree,
so, it simplifies the what group for sysfs aliases;
- the logic which splits regular expressions into subgroups
was re-written to make it faster;
- it may optionally use multiple processes to search for symbol
matches;
- it has some additional debug levels.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Jonathan Corbet <[email protected]>
Link: https://lore.kernel.org/r/1529c255845d117696d5af57d8dc05554663afdf.1739182025.git.mchehab+huawei@kernel.org

show more ...

docs: sphinx/kernel_abi: reduce buffer usage for ABI messages

Instead of producing a big message with all ABI contents and then
parse as a whole, simplify the code by handling each ABI symbol
in sep

docs: sphinx/kernel_abi: reduce buffer usage for ABI messages

Instead of producing a big message with all ABI contents and then
parse as a whole, simplify the code by handling each ABI symbol
in separate. As an additional benefit, there's no need to place
file/line nubers inlined at the data and use a regex to convert
them.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Jonathan Corbet <[email protected]>
Link: https://lore.kernel.org/r/15be22955e3c6df49d7256c8fd24f62b397ad0ff.1739182025.git.mchehab+huawei@kernel.org

show more ...

scripts/lib/abi/abi_parser.py: use an interactor for ReST output

Instead of printing all results line per line, use an interactor
to return each variable as a separate message.

This won't change mu

scripts/lib/abi/abi_parser.py: use an interactor for ReST output

Instead of printing all results line per line, use an interactor
to return each variable as a separate message.

This won't change much when using it via command line, but it
will help Sphinx integration by providing an interactor that
could be used there to handle ABI symbol by symbol.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Jonathan Corbet <[email protected]>
Link: https://lore.kernel.org/r/e3c94b8cdfd5e955aa19a703921f364a89089634.1739182025.git.mchehab+huawei@kernel.org

show more ...

scripts/get_abi.py: add support for symbol search

Add support for searching symbols from Documentation/ABI using
regular expressions to match the symbols' names.

Signed-off-by: Mauro Carvalho Cheha

scripts/get_abi.py: add support for symbol search

Add support for searching symbols from Documentation/ABI using
regular expressions to match the symbols' names.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Jonathan Corbet <[email protected]>
Link: https://lore.kernel.org/r/21b2c48657dde112d5417dcd7e0aa7cd383b9a0a.1739182025.git.mchehab+huawei@kernel.org

show more ...

scripts/get_abi.py: add a Python tool to generate ReST output

The get_abi.pl script is requiring some care, but it seems that
the number of changes on it since when I originally wrote it
was not too

scripts/get_abi.py: add a Python tool to generate ReST output

The get_abi.pl script is requiring some care, but it seems that
the number of changes on it since when I originally wrote it
was not too high.

Maintaining perl scripts without using classes requires a higher
efforted than on python, due to global variables management.
Also, it sounds easier to find python developer those days than
perl ones.

As a plus, using a Python class to handle ABI allows a better
integration with Sphinx extensions, allowing, for instance,
to let automarkup to generate cross-references for ABI
symbols.

With that in mind, rewrite the core of get_abi.pl in Python,
using classes, to help producing documentation. This will
allow a better integration in the future with the Sphinx
ABI extension.

The algorithms used there are the same as the ones in Perl,
with a couple of cleanups to remove redundant variables and
to help with cross-reference generation. While doing that,
remove some code that were important in the past, where
ABI files weren't using ReST format.

Some minor improvements were added like using a fixed seed
when generating ABI keys for duplicated names, making its
results reproductible.

The end script is a little bit faster than the original one
(tested on a machine with ssd disks). That's probably because
we're now using only pre-compiled regular expressions, and it
is using string replacement methods instead of regex where
possible.

The new version is a little bit more conservative when
converting text to cross-references to avoid adding them into
literal blocks.

To ensure that the ReST output is parsing all variables
and files properly, the end result was compared using diff
with the one produced by the perl script and showed no regressions.
There are minor improvements at the results, as it now
properly groups What on some special cases. It also better
escape some XREF names.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Jonathan Corbet <[email protected]>
Link: https://lore.kernel.org/r/71a894211a8b69664711144d9c4f8a0e73d1ae3c.1739182025.git.mchehab+huawei@kernel.org

show more ...