for getexecpath() in execve(), the addition of REALPATH to the namei LOOKUP
operation can now fail if the buffer generated becoes too long. What we want
is for the LOOKUP to succeed and and for the REALPATH 'request' to be
abandoned. Then the LOOKUP attempt can stil proceed, and execve() can
succeed at running the binary (and not provide getexecpath information).
This creates a seperate identifiable EXECPATH request which can be abandoned
easier.
issue found by dgl, change ok dgl
avoid waking up when closing the lid after suspending with machdep.lidaction=0
we do not need to check if lid_action is non-zero because we know that
we are resuming and can rely on acpibtn_numopenlids() to check if the
lid is actually open
ok kettenis@
httpd: enforce "max request body" for chunked requests passed to fastcgi
The limit was only applied to requests with a Content-Length header.
With "pass chunked" (off by default), a chunked request bypassed it,
so a client could send a body of any size to the backend.
Count the bytes of all chunks and answer 413 once they exceed the limit.
Found and provided patch by Acts1631, thanks
OK kirill@
+7zip -p7zip, add quirk, change depends
Patch x11/gnome/file-roller and x11/mate/engrampa, because 7zip does
not have p7zip's -l flag. Dropping -l fixes adding a file to a
.tar.7z archive.
Patch x11/xarchiver to find 7zCon.sfx in the new spot.
ok sthen@
Add a test where the executable does not use substitutions but dlopen()'s
a shared lib which depends on another lib found via the rpath of the
shared lib which uses ORIGIN, OSNAME, OSREL & PLATFORM.
For prog3/libyy, move the location of libaa.so to be under $ORIGIN/
OpenBSD/`uname -r`/`uname -m`/ to futher test OSNAME, OSREL, PLATFORM
substitution.
