MCEdit: incorrect syntax detection

[andrew_b]

From ticket:4246#comment:6

There are two conditions are used to detect syntax highlighting type:

• a file name matched to the specified regular expression;
• a first line of file matched to the specified regular expression (for example, a shebang in shell scripts).

Current implementation is an OR logic: if one of those conditions is met, the rule is considered found (see src/editor/syntax.c).

   1327             q = mc_search (args[1], DEFAULT_CHARSET, editor_file, MC_SEARCH_T_REGEX);
   1328             /* does filename match arg 1 ? */
   1329             if (!q && args[3] != NULL)
   1330             {
   1331                 /* does first line match arg 3 ? */
   1332                 q = mc_search (args[3], DEFAULT_CHARSET, first_line, MC_SEARCH_T_REGEX);
   1333             }
   1334             if (q)
   1335             {
   1336                 int line_error;
   1337                 char *syntax_type;
   1338 
   1339               found_type:
   1340                 syntax_type = args[2];

It is suggested to use AND logic: both regexps must be matched (if the first line regexp is present, of course) to detect the syntax highlighting type:

src/editor/syntax.c

@@ -1326 +1326 @@
 
             q = mc_search (args[1], DEFAULT_CHARSET, editor_file, MC_SEARCH_T_REGEX);
             /* does filename match arg 1 ? */
-            if (!q && args[3] != NULL)
+            if (q && args[3] != NULL)
             {
                 /* does first line match arg 3 ? */
                 q = mc_search (args[3], DEFAULT_CHARSET, first_line, MC_SEARCH_T_REGEX);

[ossi]

you can't make it AND, because that would defeat the idea that scripts without a file extension are detected solely by their shebang.

if openrc scripts have no extension "by definition", then their file name pattern must not match anything - either via a special value which makes it explicit, or by an arbitrary "impossible" file name, say n0eXt3ns10n (yeah, that looks stupid).

[andrew_b]

Replying to ossi:

you can't make it AND, because that would defeat the idea that scripts without a file extension are detected solely by their shebang.

This is an argument.

if openrc scripts have no extension "by definition", then their file name pattern must not match anything - either via a special value which makes it explicit, or by an arbitrary "impossible" file name, say n0eXt3ns10n (yeah, that looks stupid).

What about creating of a random file name, like

dd if=/dev/urandom bs=64 count=1 2>/dev/null | md5sum | cut -d' ' -f 1

[andrew_b]

Replying to andrew_b:

What about creating of a random file name

Or use ^$ pattern as it was in the first version of patch (ticket:4246#comment:1).

[ossi]

the random pattern would be just as cryptic, even if a bit less stupid-looking.
yeah, the empty pattern does seem like a good idea - after all, the actual file name can't be empty by definition.

[andrew_b]

ticket:4246#comment:9