Closed p5pRT closed 21 years ago
I use Apache Web Server 1.3.9 on Windows 98 and in the error.log I can read this:
[Mon Mar 13 14:10:32 2000] [error] [client 192.168.1.3] Premature end of script headers: c:/program files/apache group/apache/cgi-bin/acctman.pl [Mon Mar 13 14:10:32 2000] [error] [client 192.168.1.3] flock() unimplemented on this platform at c:/program files/apache group/apache/cgi-bin/acctman.pl line 672
I'm not programmer but if you can help me\, it's good Thank you
At 12:49 +0100 2000-03-13\, RATSIMANDRESY Arison wrote:
flock() unimplemented on this platform at c:/program files/apache group/apache/cgi-bin/acctman.pl line 672
...
Summary of my perl5 (5.0 patchlevel 5 subversion 03) configuration: Platform: osname=MSWin32\, osvers=4.0\,
flock() is an advisory file-locking mechanism which allows cooperating programs to share files without fighting over them. The file README.win32 distributed with perl 5.005_03 (and the forthcoming 5.6.0) says
* The following functions are currently unavailable: `fork()'\, `dump()'\, `chown()'\, `link()'\, `symlink()'\, `chroot()'\, `setpgrp()' and related security functions\, `setpriority()'\, `getpriority()'\, `syscall()'\, `fcntl()'\, `getpw*()'\, `msg*()'\, `shm*()'\, `sem*()'\, `alarm()'\, `socketpair()'\, `*netent()'\, `*protoent()'\, `*servent()'\, `*hostent()'\, `getnetby*()'. This list is possibly incomplete.
I don't know whether flock() should be a member of that list or not. Can somebody who knows more about Perl on Windows help\, please?
Whether or not Perl's documentation needs updating\, Arison\, you may get more immediate help by taking the query to the newsgroup news:comp.infosystems.www.authoring.cgi. Good luck. -- Dominic Dunlop
On Mon\, 13 Mar 2000 14:53:30 +0100\, Dominic Dunlop wrote:
flock() is an advisory file-locking mechanism which allows cooperating programs to share files without fighting over them. The file README.win32 distributed with perl 5.005_03 (and the forthcoming 5.6.0) says
\* The following functions are currently unavailable​: \`fork\(\)'\, \`dump\(\)'\, \`chown\(\)'\, \`link\(\)'\, \`symlink\(\)'\, \`chroot\(\)'\, \`setpgrp\(\)' and related security functions\, \`setpriority\(\)'\, \`getpriority\(\)'\, \`syscall\(\)'\, \`fcntl\(\)'\, \`getpw\*\(\)'\, \`msg\*\(\)'\, \`shm\*\(\)'\, \`sem\*\(\)'\, \`alarm\(\)'\, \`socketpair\(\)'\, \`\*netent\(\)'\, \`\*protoent\(\)'\, \`\*servent\(\)'\, \`\*hostent\(\)'\, \`getnetby\*\(\)'\. This list is possibly incomplete\.
I don't know whether flock() should be a member of that list or not. Can somebody who knows more about Perl on Windows help\, please?
That list from README.win32/perlwin32.pod should be removed--the canonical list can be found in perlport.pod (which is far more useful for comparing portability\, since it includes all other systems as well).
Sarathy gsar@ActiveState.com
At 09:37 -0800 2000-03-13\, Gurusamy Sarathy wrote:
That list from README.win32/perlwin32.pod should be removed--the canonical list can be found in perlport.pod (which is far more useful for comparing portability\, since it includes all other systems as well).
Well\, perlport suggests that README.win32 is (more) definitive:
The list may well be incomplete\, or even wrong in some places. When in doubt\, consult the platform-specific README files in the Perl source distribution\, and any other documentation resources accompanying a given port.
However\, at the risk of sending people round in circles\, I append a patch (^Ms lost). I've taken the opportunity to update web links (on the assumption that www.cpan.org is now the official starting point for CPAN access)\, but don't have a replacement for the now-dead ftp://fractal.mta.ca/pub/crypto/SSLeay/DES/.
There's more work to be done by somebody who knows more about the port than I do. For example\, there's no mention whatever in the README of Windows 98\, never mind 2000; and the notes on globbing need attention now that perl does it without an external helper. (This is true of perlport.pod as well.)
Anyway\, that patch:
On Tue\, 14 Mar 2000\, Dominic Dunlop wrote:
- http://cpan.perl.org/authors/id/GSAR/dmake-4.1pl1-win32.zip + http://www.cpan.org/authors/Gurusamy_Sarathy/dmake-4.1pl1-win32.zip
- http://www.perl.com/CPAN/authors/id/NI-S/Make-0.03.tar.gz + http://www.cpan.org/authors/Nick_Ing-Simmons/Make-1.00.tar.gz
- http://www.perl.com/CPAN/authors/id/GSAR/libwin32-0.14.zip + http://www.cpan.org/authors/Gurusamy_Sarathy/libwin32-0.151.zip
Isn't the canonical path for CPAN directories the authors/id/ version? Especially since the "verbose" links in authors/ are no longer maintained\, and http://www.cpan.org/authors/00.Directory.Is.Not.Maintained.Anymore says that the directory is expected to go away in the future (except for the subdirectory id).
The same file mentioned above also says that "every click on this directory entry costs a lot of time" and that "they cannot be removed because lots of people have made links to individual entries here.
Cheers\, Philip -- Philip Newton \newton@​newton\.digitalspace\.net
The same file mentioned above also says that "every click on this directory entry costs a lot of time"
I find that a bit silly. Apparently the writer and I have different ideas of "a lot".
--tom
On Tue\, 14 Mar 2000\, Tom Christiansen wrote:
The same file mentioned above also says that "every click on this directory entry costs a lot of time"
I find that a bit silly. Apparently the writer and I have different ideas of "a lot".
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Cheers\, Philip -- Philip Newton \newton@​newton\.digitalspace\.net
The same file mentioned above also says that "every click on this directory entry costs a lot of time"
I find that a bit silly. Apparently the writer and I have different ideas of "a lot".
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Every click is sacred. Every click is great. If a click is wasted\, Root gets quite irate.
--tom
Philip Newton \newton@​newton\.digitalspace\.net writes:
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Think "single large directory" and note that the authors/id space is being restructured into subdirectories.
-- Russ Allbery (rra@stanford.edu) \<http://www.eyrie.org/~eagle/>
Philip Newton \newton@​newton\.digitalspace\.net writes:
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Think "single large directory" and note that the authors/id space is being restructured into subdirectories.
namei() is a very busy beaver on big directories. And on Linux filesystems\, it takes O(n) time\, although I believe on BSD\, it takes like O(log n) time. If that's really it\, then that's why find and whatever -R take 10x the time on Linux for recurses down through big fat directories compared with the same on BSD. They make up for it in other ways\, occasionally\, like with asynchronous directory updates--well\, if you can call that a feature; fsck sometimes does not. :-(
--tom
On 14 Mar 2000\, Russ Allbery wrote:
Philip Newton \newton@​newton\.digitalspace\.net writes:
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Think "single large directory" and note that the authors/id space is being restructured into subdirectories.
Because each of the single-letter subdirectories was getting too big?
Cheers\, Philip -- Philip Newton \newton@​newton\.digitalspace\.net
On Tue\, Mar 14\, 2000 at 07:15:36AM -0800\, Russ Allbery wrote:
Philip Newton \newton@​newton\.digitalspace\.net writes:
I\, too\, wondered about that. How does accessing those directories cost extra time? If I understand correctly\, it's just a symbolic link to traverse\, which shouldn't be that costly.
Think "single large directory" and note that the authors/id space is being restructured into subdirectories.
Yes. Tom would probably know that as the "indirect inode" problem that most unix systems have with very large directories. The authors directory just won't scale. Thats why new PAUSE authors only get directories like authors/id/F/FO/FOO.
Tim.
On Tue\, 14 Mar 2000 11:05:34 +0100\, Dominic Dunlop wrote:
There's more work to be done by somebody who knows more about the port than I do. For example\, there's no mention whatever in the README of Windows 98\, never mind 2000; and the notes on globbing need attention now that perl does it without an external helper. (This is true of perlport.pod as well.)
Thanks. This should be a little better.
Sarathy gsar@ActiveState.com
Migrated from rt.perl.org#2357 (status was 'resolved')
Searchable as RT2357$