Age | Commit message (Collapse) | Author | |
---|---|---|---|
2001-11-23 | Revert change to Popen() to only wrap command in a shell if it | Todd C. Miller | |
contains meta characters since people may expect their aliases to work. Instead, just remove the "kill(lock_pid, SIGTERM);" since lockspool now does the right thing when the pipe is closed. | |||
2001-11-22 | Remove broken interrupt handler. | Marc Espie | |
Nothing of what it did is allowed in an interrupt handler. Rather, just mark that an interrupt occurred, and check for interrupts in the wait of RunCommand (that's the correct point to check for interrupts anyways). Okay miod@, millert@ approves (as he's too tired to look at the code) | |||
2001-11-22 | Explicitly mark nodes whose commands have been filled with | Marc Espie | |
implicit (suffix) rules. Then, only expand the IMPSRC/< variable if the node has been marked. This matches what Single Unix 2 and common sense say: implicit rules shouldn't count when an explicit rule has been found (an explicit rule being a full-scale dependency, with some associated commands) Note that Single Unix leaves the `PREFIX' question open, so we leave the PREFIX code as it is. This fixes regression case mk14, which now fails as it should. This is just a bug-fix. Some more correct (and faster) code should probably be substituted. Namely, right now, the suffix code is too greedy, whereas it should test for explicit rules earlier, and not even bother instantiating implicit rules from templates when they duplicate actual existing rules. ok millert@ | |||
2001-11-22 | Bad .IMPSRC. | Marc Espie | |
2001-11-22 | volatile sig_atomic_t | Markus Friedl | |
2001-11-22 | o) fix return value; | Mike Pechkin | |
o) initial KNF; millert@ help & ok; | |||
2001-11-22 | don't use "\n" in fatal() | Kevin Steves | |
2001-11-21 | Fix the fix. Thanks obj/ | Marc Espie | |
2001-11-21 | Not an implicit source. If it's ever used except to check the dependency, | Marc Espie | |
you would get bitten. | |||
2001-11-21 | o kill strcpy() | Todd C. Miller | |
o check return values of malloc and friends o use strdup() when sensible | |||
2001-11-21 | more on passphrase construction; ok markus@ | Kevin Steves | |
2001-11-21 | fix arg reversal that was introduced during ANSIfication | Todd C. Miller | |
2001-11-21 | mem leak | Markus Friedl | |
2001-11-21 | Fix extern decl of version missed in previous commit. | Todd C. Miller | |
2001-11-21 | o ANSIfy | Todd C. Miller | |
o Style nits o Use const to silent stupid -Wall warnings o strnc{py,at} -> strlc{py,at} o Use strpbrk() instead of homegrown anyof() o Use NULL instead of #defines with 0 cast to a pointer This still could use a proper audit | |||
2001-11-20 | In private version of popen(), only wrap a command to be run in a | Todd C. Miller | |
shell if it contains meta chars. Sneaky hack to work around a ksh bug. | |||
2001-11-20 | Major signal overhaul. We no longer longjmp all over the place. | Todd C. Miller | |
Instead, routines responsible to gathering user input (or in some cases outputting data) catch the signals and set flags as needed. Because of this some handlers are install without the SA_RESTART flag so syscalls are not restarted and we can check the flag. All signal handlers are now safe. This should make the flow of control a bit more grokable but the code is still ugly. | |||
2001-11-19 | more than typo | Theo de Raadt | |
2001-11-19 | kill more registers | Mike Pechkin | |
millert@ ok | |||
2001-11-19 | clarify/state that private keys are not exposed to clients using the agent; | Kevin Steves | |
ok markus@ | |||
2001-11-19 | fd leak on HUP; ok stevesk@ | Markus Friedl | |
2001-11-19 | guess what! more signal race fixes... | Theo de Raadt | |
2001-11-18 | slightly expand the Csw/Trp/Sys/Int/Sof spacing | Theo de Raadt | |
2001-11-18 | After a simplistic rewrite, signal flags were not being noticed until a | Theo de Raadt | |
succesfull read operation. To avoid this, we must use SA_RESTART, and rewrite most of the guts to handle interrupted read, write, etc, even being careful to check what library routines are called.... input from millert, markus, and espie | |||
2001-11-18 | spelling | Theo de Raadt | |
2001-11-17 | volatile sig_atomic_t | Theo de Raadt | |
2001-11-17 | Repair brain-damaged struct sigaction initialization. | Todd C. Miller | |
Noticed by Theo. | |||
2001-11-17 | o Block signals when running fts_read() so the SIGINFO handler will | Todd C. Miller | |
be sure to have a valid "entry" pointer o In SIGINFO handler, do nothing if "entry" is NULL (Theo) | |||
2001-11-17 | do not exit() from signal handler | Theo de Raadt | |
2001-11-17 | errno saving | Theo de Raadt | |
2001-11-17 | enum/int type cleanup where it made sense to do so; ok markus@ | Kevin Steves | |
2001-11-17 | properly bail out of incfile(); millert ok | Theo de Raadt | |
2001-11-17 | make signal handler safe, by blocking signals around the things that it ↵ | Theo de Raadt | |
cares about; millert ok | |||
2001-11-16 | signal handler is now safe, does not end up calling exit() | Theo de Raadt | |
2001-11-16 | errno safety in signal handler | Theo de Raadt | |
2001-11-16 | incorrect write | Theo de Raadt | |
2001-11-16 | errno safety in SIGCHLD handler | Theo de Raadt | |
2001-11-16 | Instead of using a longjmp to catch SIGPIPE, just set to SIG_IGN and | Todd C. Miller | |
check the return value on writes for error. Save and restore terminal modes when piping to a command so we end up with a known good state if the command terminates uncleanly. | |||
2001-11-16 | handle empty lines instead of dumping core; report from sha@sha-1.net | Markus Friedl | |
2001-11-14 | errno saving wrapping in a signal handler | Theo de Raadt | |
2001-11-14 | get rid of signal race | Theo de Raadt | |
2001-11-14 | o Don't accept URLs with user:pass@ and an extra unencoded @. | Heikki Korpela | |
@ should be encoded as %40 for passwords and usernames. @ in the pathname portion of the URL is ok. o decode usernames and passwords as per FTP URL specification. spotted by krw@, input from krw@ and fgs@ millert@ ok | |||
2001-11-13 | enter 3.0.1 | Markus Friedl | |
2001-11-12 | enable authorized_keys2 again. tested by fries@ | Markus Friedl | |
2001-11-12 | upgrade .su railways another hundred years; dfa@ ok | Michael Shalayeff | |
2001-11-12 | fix krb5 authorization check. found by <jhawk@MIT.EDU>. from art@, deraadt@ ok | Dug Song | |
2001-11-11 | make AuthorizedKeysFile2 fallback to AuthorizedKeysFile if ↵ | Markus Friedl | |
AuthorizedKeysFile is specified. | |||
2001-11-11 | Fixed version... don't see how this could work on i386, since it didn't | Marc Espie | |
initialize create in main.c. | |||
2001-11-11 | undo changes that crash on (at least) the alpha | Theo de Raadt | |
2001-11-11 | Add 'S' to interactive mode to toggle the display of system | Federico G. Schwindt | |
processes; From NetBSD. millert@ ok. |