| Age | Commit message (Collapse) | Author |
|---|
|
Catch up with the implementation of the "parallel" decoder, which now
flushes previously accumulated information when the end of input data
is seen. Which completes the last "item" (data pattern on the bus), and
may complete another "word" (data value that spans several bus cycles).
|
|
The pdtest(1) utility interprets test.conf files. Extend the syntax of
'input' lines. These used to accept one filename only. Add support for
'format' and 'option' keywords. These translate to runtc(1) -I options.
|
|
Introduce the -I command line option for the tstrun(1) utility, accept
input file format names, and options for the input module. Take single
words in multiple -I specs to simplify the internal interface between
the pdtest(1) and the tstrun(1) tools. This follows the existing pattern
for decoder options and channel assignments. The first spec is the file
format name (use 'match' for automatic detection). Subsequent specs are
input module options.
This makes all libsigrok supported input file formats available to the
sigrok-test environment, which formerly was constrained to .sr archives
only. Although the runtc(1) utility resides in the "decoder/" hierarchy,
it can both use VCD and other files to stimulate decoder tests, as well
as cover file formats and their specific features in future tests by
checking that the imported file's content is seen or that their options
take effect.
|
|
The previous implementation expected a rather specific sequence of
SR_DF_HEADER with samplerate, then SR_DF_LOGIC. This worked with .sr
formatted session files. But does not work with other input formats.
Also process SR_DF_META packets. Pass any samplerate that we learn about
to decoders' .metadata() method. Defer decoders' .start() until the first
logic samples are seen. This improves robustness of the decoder tests.
|
|
The 'class' identifier is a reserved word in the C++ language, while
development tools may combine C and C++ support and get confused. Avoid
conflicts, rename the variable to 'class_' in runtc.c source code.
|
|
Communicate EOF to the decoder session (when the library supports it).
|
|
The upstream IRMP project's decoder core is only slowly moving, we can
consider the IRMP decoder integration to be stable and don't expect the
output to change violently any longer. The issue of requiring a single
core instance remains and affects the GUI, but not the single threaded
test suite.
This set of test cases re-uses the NEC, RC5, RC6, and SIRC dumps which
are covered by individual decoders, too. Ideally detection results would
be identical, but in practise the annotation positions and the level of
details will differ between implementations due to their internal
operation and design choices.
The IRMP test set also covers dumps which are not covered by other IR
decoders. It's interesting to see how not all key repetitions are caught
and how single press might be missed as well. It's valuable to remain
aware during maintenance, and see how occassional failure changes. This
is why an empty output for a non-empty dump is kept here, too.
|
|
Signed-off-by: Ben Gardiner <ben.l.gardiner@gmail.com>
|
|
|
|
Python ships with a difflib(3) module which is used in the pdtest(1)
utility. The Differ.compare() routine is rather expensive, especially
when the set of input text becomes "large" (a few thousand lines). Use
the less expensive unified_diff() routine instead which does not suffer
from that cost. From the sigrok-test perspective the resulting data is
as usable.
|
|
Catch up with the parallel decoder's implementation. Documented Python
output includes the bus width and cycle count for items and words.
[ see a word-diff for the commit's essence ]
|
|
The parallel decoder can also be used to interpret SQI communication.
Using the chip select as the reset/enable signal increases robustness.
Implement a test case to cover this recently added decoder feature.
|
|
Catch up with the "parallel" decoder's implementation. Annotations for
demultiplexed words are sent after all their individual bits were seen.
Data content remains identical, this is exclusively about emission order.
|
|
The decoder implementation changed the STOP width from 0.652ms to 0.562ms,
which affects the STOP bit and the REMOTE button code annotations' end.
|
|
The decoder implementation got extended to support EEPROM reads. Adjust
the test expectation.
|
|
The decoder implementation moved the "Unknown command" message from the
PE annotation class (programming enable instruction) to warnings. Adjust
the test expectation.
|
|
Add a simple test on a short snippet of the SDQ capture. This covers the
extraction of bits and bytes, and the synchronization on BREAK symbols.
|
|
[ gsi: rephrased commit message, trimmed down capture for test ]
|
|
The CAN decoder implementation got extended to output Python data.
Adjust the test expectation to cover these on some dump files.
|
|
A capture with extended NEC protocol data became available. Add tests
for failed interpretation with the standard approach, and successful
interpretation when the extended protocol gets selected.
|
|
|
|
|
|
|
|
yet dummy, as I could not execute the test myself.
Test data (ws281x_RGBW_4mhz_snippet.sr) is already available.
|
|
Add tests for the caliper decoder. Cover positive and negative numbers,
automatic millimeter and inch detection, and omitted annotations when
values haven't changed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This commit adjusts the test configuration, but does not update the test
expectation. This remains for another commit.
|
|
The capture which was attached to bug #1460 demonstrated a previously
unhandled condition, the absence of a host activity after the data byte
transmission. Add another test case after the decoder implementation got
adjusted.
|
|
Catch up with a recent change in the PS/2 decoder implementation. End
the STOP bit at the rising CLOCK edge already, don't expect a clock
inhibit phase after a data byte got transmitted.
|
|
|
|
|
|
|
|
|
|
|
