gsub does not handle \ consistently

[Schplurtz le déboulonné]

Binary package hint: mawk

Hello,

$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 6.06.1 LTS
Release: 6.06
Codename: dapper

$ dpkg -l mawk
ii mawk 1.3.3-11ubuntu1 a pattern scanning

The interpretation of "\\" in replacement is semi random and does not conform to the man page description.

                    echo '\' | mawk '{ gsub( /\\/, "\\ \\" ) }'
produces the expected and documented answer
                    \ \
but
                    echo '\' | mawk '{ gsub( /\\/, "\\\\" ) }'
only produce one single backslash instead of two. Others awks (gawk, Solaris nawk, MacOSX 10.4.8 awk) do print 2 backslashes.

/Christophe

[Daniel T Chen (crimsun)]

Is this symptom still reproducible in 8.10 beta? (Note that the syntax seems ill on bash, mksh, ...)

[Peter Cordes (peter-cordes)]

Ubuntu AMD64 8.10.

peter@tesla:~$ echo '\' | mawk '{ gsub( /\\/, "\\\\" ); print }'
\
peter@tesla:~$ echo '\' | mawk '{ gsub( /\\/, "\\ \\" ); print }'
\ \

The syntax is fine, it's all single-quoted. The submitter left out a print, though. You can check that mawk is getting the right args and input with
strace -eexecve,read,write mawk
That produces a veritable forest of backslashes, so you can definitely see that they're not being eaten by bash.

There seem to be a few bugs in mawk's gsub: e.g.
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=100808

 This bug is a different kind of wrong matching of regexes, so I didn't mark this bug as being the same bug, though.

[rusivi2 (rusivi2-deactivatedaccount)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. I noticed the package your bugging is updated in Lucid. Does this occur in it? If so, please feel free to mark this bug as new. Thanks in advance!

[Thomas Dickey (dickey-his)]

On Wed, 15 Sep 2010, rusivi1 wrote:

> Thank you for taking the time to report this bug and helping to make
> Ubuntu better. I noticed the package your bugging is updated in Lucid.
> Does this occur in it? If so, please feel free to mark this bug as new.
> Thanks in advance!

It is a bug unrelated to changes in upstream (has to do with the way
tokens are stored). It's not addressed in any of the package changes.

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net

[Schplurtz le déboulonné (schplurtz)]

Hello,

The bug is still there in Lucid.

After reading Thomas Dickey's comment, I think there is no need to mark the bug as new. Am I right ?

/Christophe

[Thomas Dickey (dickey-his)]

On Fri, 17 Sep 2010, Schplurtz le déboulonné wrote:

> Hello,
>
> The bug is still there in Lucid.
>
> After reading Thomas Dickey's comment, I think there is no need to mark
> the bug as new. Am I right ?

right - it's a longstanding bug. iirc, mawk is reading the token in a
context where it's not quite sure that it's a pattern.

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net