[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Issue with ls -v / sort -V and strverscmp() usage

From: Bruno Haible
Subject: Re: Issue with ls -v / sort -V and strverscmp() usage
Date: Wed, 17 Sep 2008 13:13:03 +0200
User-agent: KMail/1.5.4

Hello Kamil,

Kamil Dudka wrote:
> I ran a few series of tests with all implementations. I think the 
> modified version of ververcmp (from dpkg) gives the best results - consider 
> attached results of an example - results.tar.bz2

I agree, out-deb-patched matches most closely what one would expect from
file names (in particular the file names you chose for the test case).

I'm therefore in favour of adding this function to gnulib.

There are still a few things to do before we can add it to gnulib:
  - Do you have a copyright assignment for contributions to gnulib on file?
  - Under which copyright is the original verrevcmp code? If it is not owned
    by the FSF, we can quickly redevelop it, from a formal description of this
  - Write a .h file that declares the function and describes what it does.
  - Write a module description.
  - Fix the obvious bugs, for example, you need to cast 'char' values to
    'unsigned char' before passing them to <ctype.h> functions.
  - Streamline the operation: Can't you get rid of copying the two strings?
    For example, by changing verrevcmp to take 4 arguments
    (str1, len1, str2, len2) instead of NUL terminated arguments (str1, str2)?
  - Remove unnecessary casts, for example, if match_suffix took 'char *'
    arguments and returned a 'char *', no casts would be necessary.
  - GNU coding style: space between function name and opening parenthesis,
    write 'const char *', not 'const char*', etc.
  - Put the unit test in the usual form used by gnulib. See the module
    'printf-posix-tests' for a test that produces some textual output and
    compares it to an expected output.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]