tests/**/*.py.exp: protect trailing spaces with \$ - #19073

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.47%. Comparing base (1f601e8) to head (2f9da6e).
⚠️ Report is 13 commits behind head on master.

@@           Coverage Diff           @@
##           master   #19073   +/-   ##
=======================================
  Coverage   98.47%   98.47%           
=======================================
  Files         176      176           
  Lines       22811    22811           
=======================================
  Hits        22463    22463           
  Misses        348      348           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

The failing test I think is a intermittent glitch: extmod/asyncio_lock.py.

This is a little less comprehensive than I would have liked. There are trailing spaces in some .py.exp's outside of tests/cmdline and tests/feature_check:

tests/micropython/heapalloc_traceback.py.native.exp
tests/micropython/heapalloc_traceback.py.exp
tests/ports/esp32/check_err_str.py.exp
tests/ports/unix/extra_coverage.py.exp
tests/thread/thread_exc2.py.native.exp
tests/thread/thread_exc2.py.exp
tests/extmod/vfs_basic.py.exp
tests/basics/string_tstring_basic.py.exp

But those other test directories are not marked as is_special in run-tests.py. They could be. Would it make sense to make those is_special?

In addition:

tests/cmdline/repl_paste.py
tests/cmdline/repl_autocomplete_underscore.py

contain trailing spaces as input. Not so terrible, but again they're invisble

There was some CI failure in the latest run which I think will go away on a re-run (which I can't invoke without another push).

In addition:

tests/cmdline/repl_paste.py
tests/cmdline/repl_autocomplete_underscore.py

contain trailing spaces as input.

For tests/cmdline/repl_paste.py it's probably possible to use input escape sequences like {\x20} instead of a space, eg as the 4th space in the line at the end of the line.

I think tests/cmdline/repl_autocomplete_underscore.py is now fixed in this PR by removing the trailing spaces.

tests/extmod/vfs_basic.py.exp

I think it's safe to modify the test for this so it doesn't print a space at the end. Eg some logic that doesn't print dir if it's the empty string.

tests/basics/string_tstring_basic.py.exp

As above, could modify this test to not print out a space at the end of the line.

I didn't check the other .py.exp's but maybe they can all have their test tweaked so it doesn't have trailing spaces in the output. I'd prefer that over making the output special.

Ready for re-review (though I think one test is stuck).

• There are now no more bare trailing spaces in tests/ .py and .exp files. In many cases where a possibly empty string was being printed via, say, print(..., some_str), causing a trailing space, I used print(..., repr(some_str)) instead. This makes the empty string visible and prevents bare trailing spaces.

• I added ports/unix/extra_coverage.py to the tests_with_regex_output list in run-tests.py. Changing the C code to remove the trailing spaces is not easy and not worth it.

• In run-tests.py, the {\xnn} escapes were treating nn as decimal instead of hex. This only became obvious when I added some {\x20} escapes. Fixed.

• run-tests.py was not looking for .native.exp files when those files were regexp files. Fixed.

In the long run, I wonder if it would be easier to mark regexp .exp files with some special filename, like .re.exp or .expre. That would eliminate some is_special treatment and the special-case tests_with_regex_output list. But I'd leave that for another PR.

In the long run, I wonder if it would be easier to mark regexp .exp files with some special filename, like .re.exp or .expre. That would eliminate some is_special treatment and the special-case tests_with_regex_output list. But I'd leave that for another PR.

Yes, that sounds like a good idea!

@dpgeorge I think I addressed all your review remarks. One test failed in the unix qemu_arm job: thread/stress_heap.py. Again looks like some intermittent thing.

Two more tests are stuck on web fetches. I think a re-run is in order.

Rebased and merged in 6b97759 through d54e675