Re: src/usr.sbin/slowcgi: possible bug

>I recently checked the slowcgi(8) and found that it might have an issue >when buf_pos is at the end of buffer and buf_len is zero. > >Am I right? It seems that all fastcgi blocks are aligned in 8-bytes and buffer size is 8+65535+255 = 65798 bytes which is not aligned in 8-bytes. It seems that

find -delete

This option is not posix (not like that's stopped find accumulating a dozen extensions), but it is in gnu and freebsd (for 20 years). it's also somewhat popular among sysadmins and blogs, etc. and perhaps most importantly, it nicely solves one of the more troublesome caveats of find (which the man

Re: acme-client use configuration file [1 of 5]

No objection in principle.. although since some of us depend on this we might either need warning and/or a small period of overlap where the old stuff works and then we can move to the new stuff without things blowing up. On Sun, Jan 1, 2017 at 1:59 PM, Sebastian Benoit wrote:

sunxi framebuffer driver

Here is a fun little driver that supports the framebuffer that u-boot sets up for us on the sunxi platform. The mainbus(4) changes are necessary, because the framebuffer device tree node actually lives under /chosen, so we don't pick it up during the normal device tree walk. The most common

Re: if attach/detach netlocks

On Fri, Dec 30, 2016 at 18:57 +0100, Mike Belopuhov wrote: > On Thu, Dec 29, 2016 at 09:30 +0100, Martin Pieuchot wrote: > > On 29/12/16(Thu) 01:15, Alexander Bluhm wrote: > > > On Fri, Dec 23, 2016 at 12:09:32AM +0100, Martin Pieuchot wrote: > > > > On 22/12/16(Thu) 20:45, Mike Belopuhov wrote: >

Re: pflow(4) vs NET_LOCK()

On Mon, Jan 02, 2017 at 12:12:42PM +0100, Martin Pieuchot wrote: > Diff below adds an helper for pflowioctl(). The goal is to reduce the > number of error paths where the NET_LOCK() should be released. > > ok? I would not call it pflow_init(), as ..._init() functions are often called during

Re: libressl symbols

Hi, the function prototypes have been moved from he header to the .c file - you will have to revert this change as well. +1 from me for exposing the API - I recently also ran into it but didn't speak up because my use case was probably not important enough. Reyk > Am 02.01.2017 um 16:29

libressl symbols

Looks like when libcrypto was convered to use a Symbol map, some functions were not included. This change lets py-openssl be loaded by python processes. I did not include the minor version bump, as there might be more issues. Index: lib/libcrypto/Symbols.list

src/usr.sbin/slowcgi: possible bug

Hi tech@, I recently checked the slowcgi(8) and found that it might have an issue when buf_pos is at the end of buffer and buf_len is zero. Am I right? Index: slowcgi.c === RCS file: /cvs/src/usr.sbin/slowcgi/slowcgi.c,v retrieving

Re: recursive splsoftnet()

On 2 January 2017 at 12:39, Martin Pieuchot wrote: > Kill two simple recursive splsoftnet(). > > - pr_usrreq() functions are always called under NET_LOCK(). > - tcp_ident() is called by tcp_sysctl() only which asserts for the NET_LOCK() > > ok? > OK

recursive splsoftnet()

Kill two simple recursive splsoftnet(). - pr_usrreq() functions are always called under NET_LOCK(). - tcp_ident() is called by tcp_sysctl() only which asserts for the NET_LOCK() ok? Index: netinet/ip_gre.c === RCS file:

pflow(4) vs NET_LOCK()

Diff below adds an helper for pflowioctl(). The goal is to reduce the number of error paths where the NET_LOCK() should be released. ok? Index: net/if_pflow.c === RCS file: /cvs/src/sys/net/if_pflow.c,v retrieving revision 1.65

Re: acme-client use configuration file [1 of 5]

Hi Todd, I think you meant: +.Xr acme-client.conf 5 , :^) Raf On Mon, Jan 02, 2017 at 01:58:26AM GMT, Todd C. Miller wrote: > I think you also want something like the following. > > Also, acme-client.conf.5 has paths like: > domain key /etc/ssl/private/example.com.key > domain