Backslash cleanup introduced bug [284c470e 9/17]

Backslash cleanup introduced bug [284c470e 9/17]

[Wilson Snyder]

Thanks for your recent quoting cleanup commit.  One of your
cleanups broke the verilog-mode.el package regressions.
I've repaired it; no additional action is needed, but a
heads up in case this affects other packages.

This:

-       "==" "!=" "===" "!==" "<=" ">=" "==\?" "!=\?" "<->"
+       "==" "!=" "===" "!==" "<=" ">=" "==\\?" "!=\\?" "<->"

was a list of operators, not a series of regexps, so this
change will break matching `==?' (it will look for `==\?'.
Therefore the proper cleanup is:

-       "==" "!=" "===" "!==" "<=" ">=" "==\?" "!=\?" "<->"
+       "==" "!=" "===" "!==" "<=" ">=" "==?" "!=?" "<->"

Also:

 	  (verilog-string-replace-matches
-	   "\[[^0-9: \t]+\]" "" nil nil
+	   "\\[[^0-9: \t]+]" "" nil nil


While I agree it works, I think quoting literal open
brackets without quoting literal closing brackets makes it
harder to read - which closing bracket matches the literal
open bracket?

 	  (verilog-string-replace-matches
-	   "\[[^0-9: \t]+\]" "" nil nil
+	   "\\[[^0-9: \t]+\\]" "" nil nil

I find this is easier to read as the \\[ and \\] form a
matching pair.  Furthermore I can see any unquoted closing
bracket and know it is part of a character class; and see
any quoted closing bracket and know it matches a literal.

Thanks

Re: Backslash cleanup introduced bug [284c470e 9/17]

[Paul Eggert]

On 09/18/2015 05:34 AM, Wilson Snyder wrote:
> One of your
> cleanups broke the verilog-mode.el package regressions.
>

Thanks for catching that.  I think you're referring to 753fbfbe, not to 
284c470e as the Subject: line says. 284c470e was intended to not change 
behavior. 753fbfbe was intended to fix behavior, but obviously I messed 
up in that instance (and thanks for fixing it). I now see that I also 
misclassified a few changes in 284c470e, as they actually did affect 
behavior (e.g., this change in verilog-mode.el:

-                (re-search-backward "/\*")
+                (re-search-backward "/\\*")

), but I think these misclassified changes all fixed bugs so it should 
be OK.

One other thing.  In 284c470e I tried to leave bugs alone when in 
doubt.  That is, although there are undoubtedly still some bugs related 
to backslashes, when in doubt I left behavior the same as it was 
before.  For example, lisp/net/tramp-adb.el has this:

(defcustom tramp-adb-prompt
"^\\(?:[[:digit:]]*|?\\)?\\(?:[[:alnum:]\e;[]*@[[:alnum:]]*[^#\\$]*\\)?[#\\$][[:space:]]"

and both instances of \\$ should probably be just $.  I wasn't sure, 
though, so I left them alone.

Re: Backslash cleanup introduced bug [284c470e 9/17]

[Paul Eggert]

One more thing.  Re this reversion:

-       (verilog-insert " // Implicit .*\n"))
+       (verilog-insert " // Implicit .\*\n")) ;For some reason the . or 
* must be escaped...

Could you please explain why the backslash is needed before "*"?  In 
Elisp, the two strings are equal.