Commits on Feb 6, 2023

1. Fix regular expression clobbering in the lexer …

   It is a mistake for the lexer not to make a copy of regexp literals,
   because the parser may need to shift more than one of them onto its
   stack as it searches for a matching rule.
   
   For the following demonstration, the correct output is "1b":
   
           $ echo ab | ./master 'sub(/a/, "b" ~ /b/)'
           a1
   
   The debugging output confirms that /b/ clobbered /a/, incorrectly
   modifying sub's first argument:
   
           $ echo ab | ./master -d 'sub(/a/, "b" ~ /b/)' | grep reparse
           reparse <b>
           reparse <b>
   
   One of those should have been "reparse <a>".
   
   I introduced this bug in a fix [1] for memory leaks discovered by
   Todd C. Miller [2]. This commit reverts my fix and applies a slightly
   modified version of Todd's recommendation (I omit the INDEX case
   because reg_expr is stored in a node).
   
   [1] 577a67c
   [2] onetrueawk#156
   
   Testing with `valgrind --leak-check=full $awk "$prog" </dev/null`, the
   leaks produce something like this:
   
   ==473055== 4 bytes in 1 blocks are definitely lost in loss record 36 of 128
   ==473055==    at 0x484586F: malloc (vg_replace_malloc.c:381)
   ==473055==    by 0x49E54FD: strdup (strdup.c:42)
   ==473055==    by 0x409288: tostring (tran.c:522)
   ==473055==    by 0x411DF5: regexpr (lex.c:557)
   ==473055==    by 0x402E99: yyparse (awkgram.tab.c:2251)
   ==473055==    by 0x402877: main (main.c:211)
   
   Where $prog is any one of these one-liners:
   
   '/abc/; /cde/'
   '"abc" ~ /cde/; /fgh/'
   'match(/abc/, /cde/)'
   'split(/abc/, a, /cde/)'
   'sub(/abc/, /cde/)'
   'gsub(/abc/, /cde/)'
   'sub(/abc/, /cde/, a)'
   'gsub(/abc/, /cde/, a)'

   

   mpinjr committed Feb 6, 2023
   Configuration menu

   • View commit details
   • Copy full SHA for 4236e5a
   • Browse repository at this point

   Copy the full SHA

   4236e5a View commit details

   Browse the repository at this point in the history