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

Re: [leafnode-list] need help with cygwin port



Matthias Andree schrieb am 2001-05-05, 4:09:

> "Gerrit P. Haase" <gerrit.haase@xxxxxxxxxxx> writes:
> 
> > Now I start fetchnews again and it contacts the server to 
> > fetch the message ID's, it sorts out what is too old.
> > Now it tries to store the messages at disk, but it fails.
> 
> leafnode *HEAVILY* relies on several POSIX file system semantics that do
> not map onto Windows semantics. The most important that come to my mind
> right now are:
> 
> - hard links


LN(1)                          FSF                          LN(1)

NAME
       ln - make links between files

SYNOPSIS
       ln [OPTION]... TARGET [LINK_NAME]
       ln [OPTION]... TARGET... DIRECTORY

DESCRIPTION
       Create  a  link  to  the  specified  TARGET  with optional
       LINK_NAME.  If there is more than  one  TARGET,  the  last
       argument  must  be a directory;  create links in DIRECTORY
       to each TARGET.  Create hard links  by  default,  symbolic
       links  with  --symbolic.   When  creating hard links, each
       TARGET must exist.

       -b, --backup
              make a backup of each existing destination file

       -d, -F, --directory
              hard link directories (super-user only)

       -f, --force
              remove existing destination files

       -n, --no-dereference
              treat destination that is a symlink to a  directory
              as if it were a normal file

       -i, --interactive
              prompt whether to remove destinations

       -s, --symbolic
              make symbolic links instead of hard links

       -S, --suffix=SUFFIX
              override the usual backup suffix

       -v, --verbose
              print name of each file before linking

       -V, --version-control=WORD
              override the usual version control

       --help display this help and exit

       --version
              output version information and exit

       The backup suffix is ~, unless set with SIMPLE_BACKUP_SUF-
       FIX.  The version control may be set with VERSION_CONTROL,
       values are:

       t, numbered
              make numbered backups

       nil, existing
              numbered  if  numbered backups exist, simple other-
              wise

       never, simple
              always make simple backups

REPORTING BUGS
       Report bugs to <bug-fileutils@xxxxxxx>.

SEE ALSO
       The full documentation for ln is maintained as  a  Texinfo
       manual.    If  the  info  and  ln  programs  are  properly
       installed at your site, the command

              info ln

       should give you access to the complete manual.

GNU fileutils 4.0         November 1998                         1


> - atomicity of link(2)

..Hmmm, i'm missing that on my box...
 
> - atomicity of rename(2)

rename(3)                                               rename(3)

NAME
       rename - rename a file

ANSI_SYNOPSIS
       #include <stdio.h>
       int rename(const char * old, const char * new);

       int _rename_r(void * reent,
       const char * old, const char * new);

TRAD_SYNOPSIS
       #include <stdio.h>
       int rename( old, new)
       char * old;
       char * new;

       int _rename_r( reent, old, new)
       char * reent;
       char * old;
       char * new;

DESCRIPTION
       Use rename to establish a new name (the string at new) for
       a file now known by the string at old.  After a successful
       rename,  the file is no longer accessible by the string at
       old.

       If rename fails, the file named  *<[old>>  is  unaffected.
       The  conditions  for  failure depend on the host operating
       system.

       The alternate function _rename_r is a  reentrant  version.
       The  extra  argument  reent  is  a pointer to a reentrancy
       structure.

RETURNS
       The result is either 0 (when successful) or -1  (when  the
       file could not be renamed).

PORTABILITY
       ANSI C requires rename, but only specifies that the result
       on failure be nonzero. The effects of using the name of an
       existing file as *<[new>> may vary from one implementation
       to another.

       Supporting  OS  subroutines  required:  link,  unlink,  or
       rename.

SOURCE
       src/newlib/libc/stdio/rename.c

                                                                1
> - filename "suffix" does not carry a meaning or associations

No problem on cygwin. Programs are running with .exe or without.

> - availability of external access control software such as tcpd (for use
>   with inetd), or super-servers with access control such as g2s,
>   ucspi-tcp's tcpserver, xinetd (alphabetical order); for firing up
>   nntpd aka. leafnode.

I'm running 'inetd'. leafnode is started, i get the first header...
 
> Note also that VFAT stores long names very inefficiently, so the
> performance may be awful (this is a speculation however).

WinNT uses NTFS. But it will be a big problem on Win98.

> There may be more issues not covered here. You would need to review the
> entire project and plan it before you decide if you really want to port
> it.

Yep, some problems with setuid. It is handled in another way at cygwin.
I'll need to rewrite the function belonging on setuid.

> Since Windows FAT file systems and 9x/ME in general don't give a penny
> on security and access permissions, don't bother with permissions for
> Win 9x, any user can trash the entire system and see everybody else's
> files. You could consider encapsulating the entire WinNT/2k specific
> permission stuff into a separate module that can be exchanged for fake
> functions that always return success on Win 9x/ME.

Yes, i thought it will run at these systems, i tried to unset all
security on my NT-box, to emulate, but fetchnews is not writing to disk.
 
> Still, I think, porting leafnode to a non-POSIX system will require
> rewrites of major parts, and I'm wondering if rewriting a similar
> program from scratch instead would be worthwhile and less of an effort.

Nope, Cygwin is a nearly complete posix-layer, it should work.
Rewriting will be also a big deal, i think there are only few parts
to change.
 
> Did you evaluate "Hamster"? 

No, i'm not able to build it myself, because it is pascal, i don't 
like delphi/pascal.

> 
> (I'm always wondering why people even remotely consider deploying any
> server software on Win 9x/ME when there are a lot of good and easy to
> set up, localized distributions of Linux and BSD around and 486- or
> pentium-class machineswith like 8..32 MB RAM and 500...1000 MB hard
> disks are readily available second hand, often for free; not few of them
> come with leafnode as binary package, ready to install.)

I'm on WinNT 4.0 and i don't use Win98, it sucks...
NT sucks, too, but less.

I need to run my NT Server the whole time, so i'm not able to run
Linux in parallel, only way to survive is NT with cygwin on top.

I will take a look at leafnode 2.

Thanks a lot,

gph

-- 
=^..^=

-- 
leafnode-list@xxxxxxxxxxxxxxxxxxxxxxxxxxxx -- mailing list for leafnode
To unsubscribe, send mail with "unsubscribe" in the subject to the list