[BUG] parameter ... set in enclosing scope

[ratijas]

While sourcing zsh-syntax-highlighting.zsh, zsh encounter couple of warnings.

/usr/share/zsh/plugins/zsh-syntax-highlighting/zsh-syntax-highlighting.zsh:434: array parameter ZSH_HIGHLIGHT_HIGHLIGHTERS set in enclosing scope in function (anon)
_zsh_highlight_main__precmd_hook:1: array parameter _zsh_highlight_main__command_type_cache set in enclosing scope in function _zsh_highlight_main__precmd_hook

I have my WARN_CREATE_GLOBAL and WARN_NESTED_VAR zsh options turned on.

See also

#438

[ratijas]

Also, when I cloned the repository, and loaded plugin from the latest master, errors slightly relocated:

zsh-syntax-highlighting.zsh:48: scalar parameter ZSH_HIGHLIGHT_REVISION set in enclosing scope in function (anon)
zsh-syntax-highlighting.zsh:584: array parameter ZSH_HIGHLIGHT_HIGHLIGHTERS set in enclosing scope in function (anon)

[phy1729]

All of the referenced variables are declared global before being set. The test framework also sets WARN_CREATE_GLOBAL to catch issues like these. Could you see if something else in your configuration is necessary to reproduce the issue?

[ratijas]

Interesting. I'm using my dotfiles repository, branch zsh. Relevant options might be in 11.options-user.zsh.

https://github.com/ratijas/dotfiles/tree/zsh.

[ratijas]

Looking briefly at the code, I don't even know what could be wrong with it, or how zsh imagines it should be.

# Set $0 to the expected value, regardless of functionargzero.
0=${(%):-%N}
if true; then
  # $0 is reliable
  typeset -g ZSH_HIGHLIGHT_VERSION=$(<"${0:A:h}"/.version)
  typeset -g ZSH_HIGHLIGHT_REVISION=$(<"${0:A:h}"/.revision-hash)
  if [[ $ZSH_HIGHLIGHT_REVISION == \$Format:* ]]; then
    # When running from a source tree without 'make install', $ZSH_HIGHLIGHT_REVISION
    # would be set to '$Format:%H$' literally.  That's an invalid value, and obtaining
    # the valid value (via `git rev-parse HEAD`, as Makefile does) might be costly, so:
    ZSH_HIGHLIGHT_REVISION=HEAD
  fi
fi

The problem is reported at ZSH_HIGHLIGHT_REVISION=HEAD line.

[ratijas]

Oh, I'm sorry. I provided the wrong info. I have WARN_NESTED_VAR option on. Edited the top post.

On a second thought, maybe I shouldn't be using this option. It seems too restrictive without obvious benefits. What do you think?

[danielshahaf]

Oh, I'm sorry. I provided the wrong info. I have WARN_NESTED_VAR option on. Edited the top post.

The error message under master is in the second post.

Given that the option exists, z-sy-h should support it. However, for ZSH_HIGHLIGHT_REVISION there is no reason why it would be set in the enclosing scope, and for ZSH_HIGHLIGHT_HIGHLIGHTERS, no reason why it would be set but empty in the enclosing scope. Which is to say, we're in need of a reproduction recipe.

On a second thought, maybe I shouldn't be using this option. It seems too restrictive without obvious benefits. What do you think?

It seems to me that the option does have bug-finding potential. Feel free to discuss this further on zsh's support channels.

[phy1729]

I agree it seems silly to set in interactive use, but as @danielshahaf said we still ought to support it.

I can't reproduce the interactive bug, but this avoids warnings when the test harness sets WARN_NESTED_VAR

diff --git a/tests/test-highlighting.zsh b/tests/test-highlighting.zsh
index 8b564a8..cab2fe3 100755
--- a/tests/test-highlighting.zsh
+++ b/tests/test-highlighting.zsh
@@ -29,7 +29,7 @@
 # -------------------------------------------------------------------------------------------------
 
 
-setopt NO_UNSET WARN_CREATE_GLOBAL
+setopt NO_UNSET WARN_CREATE_GLOBAL WARN_NESTED_VAR
 
 # Required for add-zle-hook-widget.
 zmodload zsh/zle
@@ -85,6 +85,8 @@ else
   exit 1
 fi > >($results_filter | $root/tests/tap-colorizer.zsh)
 
+setopt NOWARN_NESTED_VAR
+
 # Overwrite _zsh_highlight_add_highlight so we get the key itself instead of the style
 _zsh_highlight_add_highlight()
 {
@@ -154,7 +156,9 @@ run_test_internal() {
     : ${CURSOR=$#BUFFER} ${PENDING=0} ${WIDGET=z-sy-h-test-harness-test-widget}
 
     # Process the data.
+    setopt WARN_NESTED_VAR
     _zsh_highlight
+    setopt NOWARN_NESTED_VAR
   }; [[ -z $RETURN ]] || return $RETURN
   unset ARG
 
diff --git a/zsh-syntax-highlighting.zsh b/zsh-syntax-highlighting.zsh
index d20dc5b..2a45136 100644
--- a/zsh-syntax-highlighting.zsh
+++ b/zsh-syntax-highlighting.zsh
@@ -45,7 +45,7 @@ if true; then
     # When running from a source tree without 'make install', $ZSH_HIGHLIGHT_REVISION
     # would be set to '$Format:%H$' literally.  That's an invalid value, and obtaining
     # the valid value (via `git rev-parse HEAD`, as Makefile does) might be costly, so:
-    ZSH_HIGHLIGHT_REVISION=HEAD
+    typeset -g ZSH_HIGHLIGHT_REVISION=HEAD
   fi
 fi
 
@@ -130,7 +130,8 @@ _zsh_highlight()
   }
 
   # Probe the memo= feature, once.
-  (( ${+zsh_highlight__memo_feature} )) || {
+  (( ${+zsh_highlight__memo_feature} )) || (){
+    setopt localoptions nowarnnestedvar
     region_highlight+=( " 0 0 fg=red, memo=zsh-syntax-highlighting" )
     case ${region_highlight[-1]} in
       ("0 0 fg=red")
@@ -175,10 +176,10 @@ _zsh_highlight()
 
   # Reset region_highlight to build it from scratch
   if (( zsh_highlight__memo_feature )); then
-    region_highlight=( "${(@)region_highlight:#*memo=zsh-syntax-highlighting*}" )
+    typeset -g region_highlight=( "${(@)region_highlight:#*memo=zsh-syntax-highlighting*}" )
   else
     # Legacy codepath.  Not very interoperable with other plugins (issue #418).
-    region_highlight=()
+    typeset -g region_highlight=()
   fi
 
   # Remove all highlighting in isearch, so that only the underlining done by zsh itself remains.
@@ -581,7 +582,7 @@ add-zsh-hook preexec _zsh_highlight_preexec_hook 2>/dev/null || {
 zmodload zsh/parameter 2>/dev/null || true
 
 # Initialize the array of active highlighters if needed.
-[[ $#ZSH_HIGHLIGHT_HIGHLIGHTERS -eq 0 ]] && ZSH_HIGHLIGHT_HIGHLIGHTERS=(main)
+[[ $#ZSH_HIGHLIGHT_HIGHLIGHTERS -eq 0 ]] && typeset -g ZSH_HIGHLIGHT_HIGHLIGHTERS=(main)
 
 if (( $+X_ZSH_HIGHLIGHT_DIRS_BLACKLIST )); then
   print >&2 'zsh-syntax-highlighting: X_ZSH_HIGHLIGHT_DIRS_BLACKLIST is deprecated. Please use ZSH_HIGHLIGHT_DIRS_BLACKLIST.'

[danielshahaf]

I can't reproduce the interactive bug, but this avoids warnings when the test harness sets WARN_NESTED_VAR

Thanks. Reviewing.

``` diff --git a/tests/test-highlighting.zsh b/tests/test-highlighting.zsh index 8b564a8..cab2fe3 100755 --- a/tests/test-highlighting.zsh +++ b/tests/test-highlighting.zsh @@ -29,7 +29,7 @@ # ------------------------------------------------------------------------------------------------- -setopt NO_UNSET WARN_CREATE_GLOBAL +setopt NO_UNSET WARN_CREATE_GLOBAL WARN_NESTED_VAR

Hmm. It's a minor issue, but it'd be useful to standardize the spelling of option names in z-sy-h (i.e., capitalization, underscores, and "no" prefix) for greppability (whether via grep(1), via eyeballing a screenful of code, via $EDITOR's "search in current file" functionalities [e.g., in Vim, «*» in normal mode — and I'm aware of «g*», but it doesn't work for finding «foo» when cursor is on «nofoo»], or via github's issues database's search). Not a blocker, of course.

# Required for add-zle-hook-widget. zmodload zsh/zle @@ -85,6 +85,8 @@ else exit 1 fi > >($results_filter | $root/tests/tap-colorizer.zsh) +setopt NOWARN_NESTED_VAR +

Why set the option for part of the initialization code and unset it here?

# Overwrite _zsh_highlight_add_highlight so we get the key itself instead of the style _zsh_highlight_add_highlight() { @@ -154,7 +156,9 @@ run_test_internal() { : ${CURSOR=$#BUFFER} ${PENDING=0} ${WIDGET=z-sy-h-test-harness-test-widget} # Process the data. + setopt WARN_NESTED_VAR _zsh_highlight + setopt NOWARN_NESTED_VAR

I wonder whether this unsetopt could break anything, e.g., if these three lines (not just the first two) are some day moved into a 'try' block and then «_zsh_highlight» croaks, the «always» block will be executed with the option remaining set. So, it feels like a try/always block is in order. (Or we could rely on LOCAL_OPTIONS being in effect.)

}; [[ -z $RETURN ]] || return $RETURN unset ARG diff --git a/zsh-syntax-highlighting.zsh b/zsh-syntax-highlighting.zsh index d20dc5b..2a45136 100644 --- a/zsh-syntax-highlighting.zsh +++ b/zsh-syntax-highlighting.zsh @@ -45,7 +45,7 @@ if true; then # When running from a source tree without 'make install', $ZSH_HIGHLIGHT_REVISION # would be set to '$Format:%H$' literally. That's an invalid value, and obtaining # the valid value (via `git rev-parse HEAD`, as Makefile does) might be costly, so: - ZSH_HIGHLIGHT_REVISION=HEAD + typeset -g ZSH_HIGHLIGHT_REVISION=HEAD

I'm not sure I'm happy with this strategy. Since the variable was already declared global above, setting WARN_NESTED_VAR requires re-declaring the variable as global on every assignment (including assignments in the same scope, augmenting assignments, and «foo=${foo:#…}» assignments, but excluding slice assignments), makes the code harder to read, violates DRY, and introduces no semantic difference; in exchange, a «typeset -g» is added that reminds the reader that the variable is global. I would consider the costs to outweigh the benefit, since the all-uppercase name already conveys that variable is global. [This paragraph is also applicable to $ZSH_HIGHLIGHT_HIGHLIGHTERS, and even to $region_highlight with s/all-uppercase/well-known/.] Incidentally, even though sourcing z-sy-h multiple times is supposed to work, the version info globals don't provide for that. Not a blocker, and low priority, of course. (With my -workers@ hat:) Would it make sense to disable WARN_NESTED_VAR at the C level for variables that are PM_SPECIAL or their name consists of uppercase ASCII and underscores only (and isn't all underscores)? (This wouldn't catch _all_ well-known variables, of course, since some of them aren't PM_SPECIAL: e.g., $REPLY, vcs_info's $hook_com.) Would it make sense for WARN_NESTED_VAR to handle augmenting assignments and slice assignments the same way? I.e., either trip on both, or trip on neither?

fi fi @@ -130,7 +130,8 @@ _zsh_highlight() } # Probe the memo= feature, once. - (( ${+zsh_highlight__memo_feature} )) || { + (( ${+zsh_highlight__memo_feature} )) || (){ + setopt localoptions nowarnnestedvar

Anonymous functions break «return»/«exit» (by design) and «"$@"» (unless the pre-5.0.7 compatibility workaround). WDYT of avoiding an anonymous function here, in order to reduce the surface area for bugs to enter through?

region_highlight+=( " 0 0 fg=red, memo=zsh-syntax-highlighting" ) case ${region_highlight[-1]} in ("0 0 fg=red") @@ -175,10 +176,10 @@ _zsh_highlight() # Reset region_highlight to build it from scratch if (( zsh_highlight__memo_feature )); then - region_highlight=( "${(@)region_highlight:#*memo=zsh-syntax-highlighting*}" ) + typeset -g region_highlight=( "${(@)region_highlight:#*memo=zsh-syntax-highlighting*}" ) else # Legacy codepath. Not very interoperable with other plugins (issue #418). - region_highlight=() + typeset -g region_highlight=() fi # Remove all highlighting in isearch, so that only the underlining done by zsh itself remains. @@ -581,7 +582,7 @@ add-zsh-hook preexec _zsh_highlight_preexec_hook 2>/dev/null || { zmodload zsh/parameter 2>/dev/null || true # Initialize the array of active highlighters if needed. -[[ $#ZSH_HIGHLIGHT_HIGHLIGHTERS -eq 0 ]] && ZSH_HIGHLIGHT_HIGHLIGHTERS=(main) +[[ $#ZSH_HIGHLIGHT_HIGHLIGHTERS -eq 0 ]] && typeset -g ZSH_HIGHLIGHT_HIGHLIGHTERS=(main)

Does the assignment on line 239 need the same fix?

…

if (( $+X_ZSH_HIGHLIGHT_DIRS_BLACKLIST )); then print >&2 'zsh-syntax-highlighting: X_ZSH_HIGHLIGHT_DIRS_BLACKLIST is deprecated. Please use ZSH_HIGHLIGHT_DIRS_BLACKLIST.' ```

[danielshahaf]

I'm not sure I'm happy with this strategy.

Which is to say: why don't we just unset the option? Cross-referencing #758.

[FranklinYu]

update

Sorry for the spam. I wasn’t aware that this error message is about warn_nested_var instead of warn_create_global. z-sy-h works perfectly under warn_create_global; I guess I’ll simply disable warn_nested_var.

original comment

With my current Zsh configuration, I can reproduce the second bug:

_zsh_highlight_main__precmd_hook:1: array parameter _zsh_highlight_main__command_type_cache set in enclosing scope in function _zsh_highlight_main__precmd_hook

This is weird since I can clearly see the variable set as global in

zsh-syntax-highlighting/highlighters/main/main-highlighter.zsh

Line 1841 in 6e0e950

typeset -gA _zsh_highlight_main__command_type_cache

I’ll try finding a minimal example to reproduce.