* src/lexer.l: properly handle backslash escaping in strings and regex

We used to look at a string literal like "\\" as not ending, and instead
mistook \" as escaping the ending quotes. This (longstanding, embarrassing)
bug would lead to weird syntax errors in lenses when encountered.

Fixes hercules-team#495

NEWS

@@ -4,6 +4,9 @@
       back to text fails. They now make it clearer what part of the tree
       was problematic, and what the tree should have looked like.
     * Fixed the pkg-config file, which should now be usable
+    * Fix handling of backslash-escaping in strings and regular expressions
+      in the lens language. We used to handle constructs like "\\" and
+      /\\\\/ incorrectly. (Issue #495)
   - API changes
     * add function aug_ns_attr to allow iterating through a nodeset
       quickly. See examples/dump.c for an example of how to use them

lenses/tests/test_httpd.aug

@@ -156,10 +156,10 @@ test Httpd.lns get c5 =
     { "arg" = "agent" }
   }
 
-let c7 = "LogFormat \"%v:%p %h %l %u %t \\"%r\\" %>s %O \\"%{Referer}i\\" \\"%{User-Agent}i\\"\" vhost_combined\n"
+let c7 = "LogFormat \"%v:%p %h %l %u %t \\\"%r\\\" %>s %O \\\"%{Referer}i\\\" \\\"%{User-Agent}i\\\"\" vhost_combined\n"
 test Httpd.lns get c7 =
   { "directive" = "LogFormat"
-    { "arg" = "\"%v:%p %h %l %u %t \\"%r\\" %>s %O \\"%{Referer}i\\" \\"%{User-Agent}i\\"\"" }
+    { "arg" = "\"%v:%p %h %l %u %t \\\"%r\\\" %>s %O \\\"%{Referer}i\\\" \\\"%{User-Agent}i\\\"\"" }
     { "arg" = "vhost_combined" }
   }
 

lenses/tests/test_json.aug

@@ -489,8 +489,8 @@ test lns get s_commented =
 
 (* Test lns
      Allow escaped quotes, backslashes and tabs/newlines *)
-test lns get "{ \"filesystem\": \"ext3\\" \\\\ \t \r\n SEC_TYPE=\\"ext2\" }\n" =
+test lns get "{ \"filesystem\": \"ext3\\\" \\\\ \t \r\n SEC_TYPE=\\\"ext2\" }\n" =
   { "dict"
     { "entry" = "filesystem"
-      { "string" = "ext3\\" \\\\ \t \r\n SEC_TYPE=\\"ext2" } }
+      { "string" = "ext3\\\" \\\\ \t \r\n SEC_TYPE=\\\"ext2" } }
     {  } }

src/lexer.l

@@ -87,20 +87,20 @@ ARROW  ->
 
 <INITIAL>
 {
-  \"([^\"]|\\\")*\"   {
+  \"([^\\\"]|\\(.|\n))*\"   {
                loc_update(yylloc, yytext, yyleng);
                yylval->string = unescape(yytext+1, yyleng-2, STR_ESCAPES);
                return DQUOTED;
   }
 
-  \/([^/]|\\\/)*\/i {
+  \/([^\\\/]|\\(.|\n))*\/i {
                loc_update(yylloc, yytext, yyleng);
                yylval->regexp.nocase = 1;
                yylval->regexp.pattern = regexp_literal(yytext+1, yyleng-3);
                return REGEXP;
   }
 
-  \/([^/]|\\\/)*\/ {
+  \/([^\\\/]|\\(.|\n))*\/ {
                loc_update(yylloc, yytext, yyleng);
                yylval->regexp.nocase = 0;
                yylval->regexp.pattern = regexp_literal(yytext+1, yyleng-2);

tests/modules/pass_lexer.aug

@@ -0,0 +1,13 @@
+module Pass_Lexer =
+
+  (* Some tests for corner cases for the lexer; they will all lead to
+   * syntax errors if we are not lexing correctly *)
+
+  let s1 = "\\"
+  let s2 = "\
+"
+
+  let r1 = /\\\\/
+
+  let slash = "/" (* Just here to cause trouble if the lexer does not
+                   * properly terminate the above expressions *)