When system calls indicate an error they return -1, not some arbitrary
value < 0. errno is only updated in this case. Change all (most?)
callers of syscalls to follow this better, and let's see if this strictness
helps us in the future.

Don't reopen file if its stdin.

From leo_tck at volny dot cz
OK millert@

Modify code added in rev 1.30 to use the correct variable instead of a
different uninitialised one.

ok martijn@

Fix a crash as found by sthen@

tested and OK sthen@
OK with minor tweak schwarze@

better fixes for running tail without -f. from Martijn van Duren

quick hack to fix "tail +n", reported broken by Mikolaj Kucharski.
this doesn't attempt to repair all cases, but restoring basic functionality
is a first step.

another try to allow tailing multiple files. maybe it works?
commit now to allow people to test.
from Martijn van Duren

rcsid[] and sccsid[] and copyright[] are essentially unmaintained (and
unmaintainable). these days, people use source. these id's do not provide
any benefit, and do hurt the small install media
(the 33,000 line diff is essentially mechanical)
ok with the idea millert, ok dms

Backout previous commit, there are still some issues with it.
ok sthen@

Permit tail -f to follow multiple files, useful when you want to monitor
several logfiles in a single terminal. split forward() and move the
corresponding -rewritten- code to follow().
Closes PR 5092.

ok millert@ gilles@ sobrado@

use EV_SET() instead of doing it by hand; Iruata Souza

unneeded includes and whitespace police; from Igor Zinovik

ANSIfy. Ok ray@

If kqueue is not available, revert back to old behavior. Unbreaks tail -f
on NFS and other filesystems not supporting kqueues.
ok tedu@ millert@

Unbreak file truncation handling. Resolves PR 3689.
ok tdeval@ millert@

make -b and -c work with large offsets; tweak from millert@
ok henning@ itojun@ millert@

typo in comment. ok henning@

rewrite to not use mmap(2). solves a bunch of issues:
- No more I/O related SEGVs (PR 2337?).
- Performance linear to size of output, not input file size, even for very large files.
Same for -r.
- Reverse tail also works for very large files.
- Very large values of n possible, even for very large files. Same for -r.

work by Otto Moerbeek <otto@drijf.net>
tested by and ok millert@ and myself

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

Part one of userland __P removal. Done with a simple regexp with some minor hand editing to make comments line up correctly. Another pass is forthcoming that handles the cases that could not be done automatically.

kill more registers

millert@ ok

Use kqueue for -f.

When doing tail -f from stdin don't reopen a local file named stdin.

MAP_FAILED, not -1

Get rid of repeating error messages; From netbsd millert@ ok

supply particular mapping type to the mmap(2) instead of bogus 0

Fix broken conversion of private err() function and make ierr() and oerr()
back into functions, not macros. Also make functions that call mmap()
return an int so we can fall back to non-mmap (slow) version on mmap failure.
Noticable effects:
o when you tail multiple files, if one of the files doesn't exist or
is otherwise broken, keep on going to the next file.
o it is now possible to tail files > 2gig, though it will be *very* slow
since tail will be unable to use mmap()

In -f mode notice when the file has been replace or truncated and
deal. Useful for tail'ing log files that get rotated.

why select(2) when you can sleep(3) cheap?

Fix core dump and use libc err(3) instead of private one (which had
slightly different semantics). Also clean up some gcc -Wall'isms.

rcsid

From NetBSD:
Don't rely on select() not to modify the timeval argument. Also,
simplify the code by passing in NULL rather than an empty fd_set to
indicate no file descriptors.

branches: 1.1.1;
Initial revision

Don't reopen file if its stdin.

From leo_tck at volny dot cz
OK millert@

Modify code added in rev 1.30 to use the correct variable instead of a
different uninitialised one.

ok martijn@

Fix a crash as found by sthen@

tested and OK sthen@
OK with minor tweak schwarze@

better fixes for running tail without -f. from Martijn van Duren

quick hack to fix "tail +n", reported broken by Mikolaj Kucharski.
this doesn't attempt to repair all cases, but restoring basic functionality
is a first step.

another try to allow tailing multiple files. maybe it works?
commit now to allow people to test.
from Martijn van Duren

rcsid[] and sccsid[] and copyright[] are essentially unmaintained (and
unmaintainable). these days, people use source. these id's do not provide
any benefit, and do hurt the small install media
(the 33,000 line diff is essentially mechanical)
ok with the idea millert, ok dms

Backout previous commit, there are still some issues with it.
ok sthen@

Permit tail -f to follow multiple files, useful when you want to monitor
several logfiles in a single terminal. split forward() and move the
corresponding -rewritten- code to follow().
Closes PR 5092.

ok millert@ gilles@ sobrado@

use EV_SET() instead of doing it by hand; Iruata Souza

unneeded includes and whitespace police; from Igor Zinovik

ANSIfy. Ok ray@

If kqueue is not available, revert back to old behavior. Unbreaks tail -f
on NFS and other filesystems not supporting kqueues.
ok tedu@ millert@

Unbreak file truncation handling. Resolves PR 3689.
ok tdeval@ millert@

make -b and -c work with large offsets; tweak from millert@
ok henning@ itojun@ millert@

typo in comment. ok henning@

rewrite to not use mmap(2). solves a bunch of issues:
- No more I/O related SEGVs (PR 2337?).
- Performance linear to size of output, not input file size, even for very large files.
Same for -r.
- Reverse tail also works for very large files.
- Very large values of n possible, even for very large files. Same for -r.

work by Otto Moerbeek <otto@drijf.net>
tested by and ok millert@ and myself

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

Part one of userland __P removal. Done with a simple regexp with some minor hand editing to make comments line up correctly. Another pass is forthcoming that handles the cases that could not be done automatically.

kill more registers

millert@ ok

Use kqueue for -f.

When doing tail -f from stdin don't reopen a local file named stdin.

MAP_FAILED, not -1

Get rid of repeating error messages; From netbsd millert@ ok

supply particular mapping type to the mmap(2) instead of bogus 0

Fix broken conversion of private err() function and make ierr() and oerr()
back into functions, not macros. Also make functions that call mmap()
return an int so we can fall back to non-mmap (slow) version on mmap failure.
Noticable effects:
o when you tail multiple files, if one of the files doesn't exist or
is otherwise broken, keep on going to the next file.
o it is now possible to tail files > 2gig, though it will be *very* slow
since tail will be unable to use mmap()

In -f mode notice when the file has been replace or truncated and
deal. Useful for tail'ing log files that get rotated.

why select(2) when you can sleep(3) cheap?

Fix core dump and use libc err(3) instead of private one (which had
slightly different semantics). Also clean up some gcc -Wall'isms.

rcsid

From NetBSD:
Don't rely on select() not to modify the timeval argument. Also,
simplify the code by passing in NULL rather than an empty fd_set to
indicate no file descriptors.

branches: 1.1.1;
Initial revision

Modify code added in rev 1.30 to use the correct variable instead of a
different uninitialised one.

ok martijn@

Fix a crash as found by sthen@

tested and OK sthen@
OK with minor tweak schwarze@

better fixes for running tail without -f. from Martijn van Duren

quick hack to fix "tail +n", reported broken by Mikolaj Kucharski.
this doesn't attempt to repair all cases, but restoring basic functionality
is a first step.

another try to allow tailing multiple files. maybe it works?
commit now to allow people to test.
from Martijn van Duren

rcsid[] and sccsid[] and copyright[] are essentially unmaintained (and
unmaintainable). these days, people use source. these id's do not provide
any benefit, and do hurt the small install media
(the 33,000 line diff is essentially mechanical)
ok with the idea millert, ok dms

Backout previous commit, there are still some issues with it.
ok sthen@

Permit tail -f to follow multiple files, useful when you want to monitor
several logfiles in a single terminal. split forward() and move the
corresponding -rewritten- code to follow().
Closes PR 5092.

ok millert@ gilles@ sobrado@

use EV_SET() instead of doing it by hand; Iruata Souza

unneeded includes and whitespace police; from Igor Zinovik

ANSIfy. Ok ray@

If kqueue is not available, revert back to old behavior. Unbreaks tail -f
on NFS and other filesystems not supporting kqueues.
ok tedu@ millert@

Unbreak file truncation handling. Resolves PR 3689.
ok tdeval@ millert@

make -b and -c work with large offsets; tweak from millert@
ok henning@ itojun@ millert@

typo in comment. ok henning@

rewrite to not use mmap(2). solves a bunch of issues:
- No more I/O related SEGVs (PR 2337?).
- Performance linear to size of output, not input file size, even for very large files.
Same for -r.
- Reverse tail also works for very large files.
- Very large values of n possible, even for very large files. Same for -r.

work by Otto Moerbeek <otto@drijf.net>
tested by and ok millert@ and myself

Remove the advertising clause in the UCB license which Berkeley
rescinded 22 July 1999. Proofed by myself and Theo.

Part one of userland __P removal. Done with a simple regexp with some minor hand editing to make comments line up correctly. Another pass is forthcoming that handles the cases that could not be done automatically.

kill more registers

millert@ ok

Use kqueue for -f.

When doing tail -f from stdin don't reopen a local file named stdin.

MAP_FAILED, not -1

Get rid of repeating error messages; From netbsd millert@ ok

supply particular mapping type to the mmap(2) instead of bogus 0

Fix broken conversion of private err() function and make ierr() and oerr()
back into functions, not macros. Also make functions that call mmap()
return an int so we can fall back to non-mmap (slow) version on mmap failure.
Noticable effects:
o when you tail multiple files, if one of the files doesn't exist or
is otherwise broken, keep on going to the next file.
o it is now possible to tail files > 2gig, though it will be *very* slow
since tail will be unable to use mmap()

In -f mode notice when the file has been replace or truncated and
deal. Useful for tail'ing log files that get rotated.

why select(2) when you can sleep(3) cheap?

Fix core dump and use libc err(3) instead of private one (which had
slightly different semantics). Also clean up some gcc -Wall'isms.

rcsid

From NetBSD:
Don't rely on select() not to modify the timeval argument. Also,
simplify the code by passing in NULL rather than an empty fd_set to
indicate no file descriptors.

branches: 1.1.1;
Initial revision