- Print the test name and data after the plan line
- Split on the plan line rather than on comments
+ That makes tap-filter more suitable to filter TAP output generated by other
TAP producers.
+ However, the filtered output deletes the plan line and adds a blank line in
its stead. This suits our use-case of interactive test runs.
In «a="b=c"; $a», the '=' sign in the expansion of $a is not active.
Therefore, prevent the expansion of $a from being considered an
assignment. Update test expectations accordingly.
As a side effect, this prevents line 836 from firing for the cases in
these two tests, thereby fixing the double $region_highlight addition
(see log message of the previous commit). That leaves the line 966
addition. However, the double addition remains a latent bug (see
discussion in PR #684).
Current test failures:
Running test main
# parameter-value-contains-command-position1
1..2
ok 1 - [1,7] «$foobar»
not ok 2 - have 1 expectations and 2 region_highlight entries: «expected_region_highlight=( '1 7 assign' )» «region_highlight=( '0 7 assign' '0 7 assign' )»
# parameter-value-contains-command-position2
1..2
ok 1 - [1,2] «$y»
not ok 2 - have 1 expectations and 2 region_highlight entries: «expected_region_highlight=( '1 2 assign' )» «region_highlight=( '0 2 assign' '0 2 assign' )»
They fail as follows:
ZSH_PATCHLEVEL=debian/5.7.1-1
Running test brackets
Running test main
# parameter-value-contains-command-position1
1..2
not ok 1 - [1,7] «$foobar» - expected (1 7 "assign"), observed (1 7 "unknown-token").
ok 2 - cardinality check
# parameter-value-contains-command-position2
1..2
not ok 1 - [1,2] «$y» - expected (1 2 "assign"), observed (1 2 "unknown-token").
ok 2 - cardinality check
(The "unknown-token" was printed by the `_zsh_highlight_main_add_region_highlight`
call on line 967, after the loop.)
Before the parent commit, they behaved as follows (note the "BUG:" and
"Bail out!" on the first one — that's issue #670.2):
ZSH_PATCHLEVEL=debian/5.7.1-1
# parameter-value-contains-command-position1
1..2
ok 1 - [1,7] «$foobar» - # TODO "issue #670"
not ok 2 - have 1 expectations and 6 region_highlight entries: «expected_region_highlight=( '1 7 assign "issue ♯670"' )» «region_highlight=( '0 7 assign' '2 7 default' '2 7 command-substitution-unquoted'
zsh-syntax-highlighting: BUG: _zsh_highlight_highlighter_main_paint: start(2) >= end(2)
Bail out! On './highlighters/main/test-data/parameter-value-contains-command-position2.zsh': output on stderr
# parameter-value-contains-command-position2
1..2
ok 1 - [1,2] «$y» - # TODO "issue #670"
ok 2 - cardinality check
Vim doesn't honour the modelines by default because they're too far from
the start/end of the file [the default value of the 'modelines' (sic)
option is too small].
We cannot now for sure whether the flag is misspelled or simply unknown to us,
so err on the side of caution. This fixes an unreleased regression. Fixes#658.
Issue #641 was originally filed about this problem, but is left open to track
further enhancements.
$last_alias isn't needed; there's no reason to treat loops of length 2
(alias a=b b=a) differently to loops of length 1 (alias a=a), length 3
(alias a=b b=c c=a), or length N.
The «(( $+seen_alias[$arg] ))» check is redundant as of the last commit:
the enclosing condition ensures that $res is "alias", which implies that
«(( $+seen_alias[$arg] ))» is false.
occurred on zsh-5.0.7 and older but I don't have zsh-5.0.7 handy to test
on.)
Evidently, the issue was due to elision.
This addresses #665.0 and #665.5.
