[BACK]Return to FIXES CVS log [TXT][DIR] Up to [local] / src / usr.bin / awk

Diff for /src/usr.bin/awk/FIXES between version 1.12 and 1.13

version 1.12, 2003/11/09 20:13:57 version 1.13, 2004/12/30 01:52:48
Line 26 
Line 26 
 This file lists all bug fixes, changes, etc., made since the AWK book  This file lists all bug fixes, changes, etc., made since the AWK book
 was sent to the printers in August, 1987.  was sent to the printers in August, 1987.
   
   Dec 22, 2004:
           cranked up size of NCHARS; coverity thinks it can be overrun with
           smaller size, and i think that's right.  added some assertions to b.c
           to catch places where it might overrun.  the RE code is still fragile.
   
   Dec 5, 2004:
           fixed a couple of overflow problems with ridiculous field numbers:
           e.g., print $(2^32-1).  thanks to ruslan ermilov, giorgos keramidas
           and david o'brien at freebsd.org for patches.  this really should
           be re-done from scratch.
   
   Nov 21, 2004:
           fixed another 25-year-old RE bug, in split.  it's another failure
           to (re-)initialize.  thanks to steve fisher for spotting this and
           providing a good test case.
   
   Nov 22, 2003:
           fixed a bug in regular expressions that dates (so help me) from 1977;
           it's been there from the beginning.  an anchored longest match that
           was longer than the number of states triggered a failure to initialize
           the machine properly.  many thanks to moinak ghosh for not only finding
           this one but for providing a fix, in some of the most mysterious
           code known to man.
   
           fixed a storage leak in call() that appears to have been there since
           1983 or so -- a function without an explicit return that assigns a
           string to a parameter leaked a Cell.  thanks to moinak ghosh for
           spotting this very subtle one.
   
   Jul 31, 2003:
           fixed, thanks to andrey chernov and ruslan ermilov, a bug in lex.c
           that mis-handled the character 255 in input.  (it was being compared
           to EOF with a signed comparison.)
   
   Jul 29, 2003:
           fixed (i think) the long-standing botch that included the beginning of
           line state ^ for RE's in the set of valid characters; this led to a
           variety of odd problems, including failure to properly match certain
           regular expressions in non-US locales.  thanks to ruslan for keeping
           at this one.
   
   Jul 28, 2003:
           n-th try at getting internationalization right, with thanks to volker
           kiefel, arnold robbins and ruslan ermilov for advice, though they
           should not be blamed for the outcome.  according to posix, "."  is the
           radix character in programs and command line arguments regardless of
           the locale; otherwise, the locale should prevail for input and output
           of numbers.  so it's intended to work that way.
   
           i have rescinded the attempt to use strcoll in expanding shorthands in
           regular expressions (cclenter).  its properties are much too
           surprising; for example [a-c] matches aAbBc in locale en_US but abBcC
           in locale fr_CA.  i can see how this might arise by implementation
           but i cannot explain it to a human user.  (this behavior can be seen
           in gawk as well; we're leaning on the same library.)
   
           the issue appears to be that strcoll is meant for sorting, where
           merging upper and lower case may make sense (though note that unix
           sort does not do this by default either).  it is not appropriate
           for regular expressions, where the goal is to match specific
           patterns of characters.  in any case, the notations [:lower:], etc.,
           are available in awk, and they are more likely to work correctly in
           most locales.
   
           a moratorium is hereby declared on internationalization changes.
           i apologize to friends and colleagues in other parts of the world.
           i would truly like to get this "right", but i don't know what
           that is, and i do not want to keep making changes until it's clear.
   
   Jul 4, 2003:
           fixed bug that permitted non-terminated RE, as in "awk /x".
   
   Jun 1, 2003:
           subtle change to split: if source is empty, number of elems
           is always 0 and the array is not set.
   
   Mar 21, 2003:
           added some parens to isblank, in another attempt to make things
           internationally portable.
   
   Mar 14, 2003:
           the internationalization changes, somewhat modified, are now
           reinstated.  in theory awk will now do character comparisons
           and case conversions in national language, but "." will always
           be the decimal point separator on input and output regardless
           of national language.  isblank(){} has an #ifndef.
   
           this no longer compiles on windows: LC_MESSAGES isn't defined
           in vc6++.
   
           fixed subtle behavior in field and record splitting: if FS is
           a single character and RS is not empty, \n is NOT a separator.
           this tortuous reading is found in the awk book; behavior now
           matches gawk and mawk.
   
 Dec 13, 2002:  Dec 13, 2002:
         for the moment, the internationalization changes of nov 29 are          for the moment, the internationalization changes of nov 29 are
         rolled back -- programs like x = 1.2 don't work in some locales,          rolled back -- programs like x = 1.2 don't work in some locales,
Legend:
Removed from v.1.12  
changed lines
  Added in v.1.13