For open/openat, if the flags parameter does not contain O_CREAT, the
3rd (variadic) mode_t parameter is irrelevant. Many developers in the past
have passed mode_t (0, 044, 0644, or such), which mig

For open/openat, if the flags parameter does not contain O_CREAT, the
3rd (variadic) mode_t parameter is irrelevant. Many developers in the past
have passed mode_t (0, 044, 0644, or such), which might lead future people
to copy this broken idiom, and perhaps even believe this parameter has some
meaning or implication or application. Delete them all.
This comes out of a conversation where tb@ noticed that a strange (but
intentional) pledge behaviour is to always knock-out high-bits from
mode_t on a number of system calls as a safety factor, and his bewilderment
that this appeared to be happening against valid modes (at least visually),
but no sorry, they are all irrelevant junk. They could all be 0xdeafbeef.
ok millert

show more ...

consider two files sharing the same inode identical

This gives a substantial speedup when comparing directory
structures with many hardlinked files, e.g. when using
rsnapshot for incremental backup.

consider two files sharing the same inode identical

This gives a substantial speedup when comparing directory
structures with many hardlinked files, e.g. when using
rsnapshot for incremental backup.

ok stsp@ millert@

show more ...

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

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.

show more ...

mkstemp() returns -1 on failure

Rectify line numbers for s/.// commands in ed-style diffs.

When diff encounters a line that consists of a single dot, it emits two
dots instead, stops the current command and emits a substitute comm

Rectify line numbers for s/.// commands in ed-style diffs.

When diff encounters a line that consists of a single dot, it emits two
dots instead, stops the current command and emits a substitute command
to replace the double dot with a single one. Then it restarts the
(original) command if necessary and inserts further lines. This is done
because a single dot on a line does have special meaning in ed. (It
stops text insertion.)

Following issues are fixed by this commit.

- The line number for the substitute command should be the number of the
most recently inserted line. diff instead uses the number of the first
inserted line of the current hunk. The first character of that line is
removed when applying the diff, while the superfluous dot is not.

- The line number of the restarted command is not adjusted for the
number of lines already inserted, resulting in the reordering of lines..

- When there is a bare dot in the replacement text of a change command,
too many lines are deleted, because a second change command is emitted.
An append command should be emitted instead, because the target lines
have already been removed by the first change command.

tweaks and ok stefan@

show more ...

remove TMPDIR support. ok millert

Replace our /^\.\././ expression with /.//. The term is simpler and has
the same meaning in our diff ed-context.

As a bonus, our ed-diff output can be processed by GNU patch now, too.

okay millert@

Remove the non-standard -l flag that pipes the output through pr(1).
Based on a diff from and OK deraadt@

xmalloc/free wrappers don't need to support 20 year old non comformance

Change internal xrealloc() to a idiom-following xreallocarray().
This loses a "new size is 0" failure case. Probably not relevant;
and since we develop this in OpenBSD, we'll catch that before someo

Change internal xrealloc() to a idiom-following xreallocarray().
This loses a "new size is 0" failure case. Probably not relevant;
and since we develop this in OpenBSD, we'll catch that before someone
else imports this...
ok millert

show more ...

Include stdint.h, not limits.h to get SIZE_MAX. OK guenther@

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_N

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

show more ...

Ensure diff -uw always produces valid output when one file doesn't end
with a newline. Issue reported by guenther@.

ok guenther@

Switch diff(1) binary file detection from !(isprint() || isspace()) to
checking for embedded NULs, as was done for grep(1) some time ago.
Avoids problems with e.g. latin1-encoded files being treated

Switch diff(1) binary file detection from !(isprint() || isspace()) to
checking for embedded NULs, as was done for grep(1) some time ago.
Avoids problems with e.g. latin1-encoded files being treated as binary, since
isprint() uses only ASCII by default and diff(1) doesn't call setlocale().
prodded by and ok bluhm

show more ...

Strip trailing slashes from directory in splice() before
appending the file portion. OK krw@ matthew@

Move an isqrt() call outside a macro, from Michael W Bombardieri.

ok ray

Close FILEs when fork fails.

OK nicm

diff exits 2 on error, not 1.

OK millert

cvs and rcs use diff_output() to print everything since it is not
always to stdout. A large number of differences are due to this.
This diff reduces many more differences between diff and cvs/rcs.
Pe

cvs and rcs use diff_output() to print everything since it is not
always to stdout. A large number of differences are due to this.
This diff reduces many more differences between diff and cvs/rcs.
Personally I think it's kinda ugly. =(

"Looks fine" millert

show more ...

Remove unused arguments from check(). From cvs/rcs.

OK millert

Return -1 on error as advertised.
Handle case when only one stream failed.

OK otto, millert

Even though this cannot happen in the diff -r case,
even though diff(1) will exit(3) soon after this mkstemp(3) failure,
even though this cannot really leak anything,
close the file descriptor as soo

Even though this cannot happen in the diff -r case,
even though diff(1) will exit(3) soon after this mkstemp(3) failure,
even though this cannot really leak anything,
close the file descriptor as soon as it is not used any more
to make correctness of the code more obvious.

patch from Igor Zinovik <zinovik dot igor at gmail dot com> on tech@
"looks nice" deraadt@

show more ...

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
(th

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

show more ...

More cvs/diff/rcs convergence:
1. Mostly variable/function renaming, SIZE_T_MAX->SIZE_MAX, and
spacing.
2. One strchr -> strncspn.
3. diff had a weird thing where it set file[12] = ofile[12] but
neve

More cvs/diff/rcs convergence:
1. Mostly variable/function renaming, SIZE_T_MAX->SIZE_MAX, and
spacing.
2. One strchr -> strncspn.
3. diff had a weird thing where it set file[12] = ofile[12] but
never updated file or ofile, then if file and ofile were different
it freed it. I removed it.

OK millert

show more ...

Pull changes from rcsdiff, similar to what was pulled into cvsdiff.

OK millert