If the slowcgi timeout fires before the command was actually started then
just abort the fastcgi request with error 408.
This can happen on POST requests where the client stops sending data.
Now bgplgd does not support POST but that is only checked once the full
fastcgi request from the webserver
On 2023/03/30 20:09:21 +0200, Theo Buehler wrote:
> > Index: main.c
> [...]
> > @@ -159,8 +162,10 @@ main(int argc, char **argv)
> > update(CMODE);
> >
> > /* user startup file. */
> > - if ((cp = startupfile(NULL, conffile)) != NULL)
> > - (void)load(cp);
> > + if (ffp) {
> I've also noticed that since startupfile() returns static memory
> errors in -u would report the wrong path (ttykeymapinit() calls
> startupfile() itself) so I'm changing it to use a caller provided
> buffer for the path. tbf i don't completely like it and it also
> causes some churn in the diff
On 2023/03/30 18:11:32 +0200, Theo Buehler wrote:
> On Thu, Mar 30, 2023 at 03:47:24PM +0200, Omar Polo wrote:
> > - else if (bufp[0] == '\0')
> > + if (bufp[0] == '\0')
> > return (FALSE);
> > - return (load(fname));
> > + if ((bufp = adjustname(fname, TRUE)) == NULL)
> > +
On Thu, Mar 30, 2023 at 04:53:05PM +0200, Otto Moerbeek wrote:
> Hi,
>
> this came up after a remark by tb@; it does a more thorough check of
> the chunks in the delay list if malloc option F (which is included in
> S).
>
> Catches way more use-after-free's as otherwise we'll have to wait
> until
On Thu, 2023-03-30 at 17:50 +0200, Omar Polo wrote:
>
> If your scripts are executable script -c ./foo (with foo being your
> script) should work. sh would exec `foo' which -hopefully- has the
> correct interpreter in its shebang (i.e. csh, sh, ksh, perl, lua...)
>
Yes, thank you.
On Thu, Mar 30, 2023 at 03:47:24PM +0200, Omar Polo wrote:
> There's this "pattern" in mg where it first access(2) a path, then
> does a buch of other things and finally opens it. It can do better.
>
> Diff below removes the access() calls for conffile handling.
> startupfile() now does an ffrope
On Thu, 30 Mar 2023 17:26:17 +0200, Omar Polo wrote:
> noticed because it renders oddly:
>
>kill %job
>kill[-s signal_name] pid
>kill -sig pid ...
>
> The Op on its own line becomes part of the item body instead of the
> item itself. Use an in-line Oo ... O
On 2023/03/30 22:59:08 +0800, lux wrote:
> On Thu, 2023-03-30 at 16:31 +0200, Omar Polo wrote:
> >
> > Furthermore, the man page explicitly states that script -c runs
> > sh(1):
> >
> > -c command
> > Run sh -c command, instead of an interactive shell.
> > [...]
> >
> > So I'd
noticed because it renders oddly:
kill %job
kill[-s signal_name] pid
kill -sig pid ...
The Op on its own line becomes part of the item body instead of the
item itself. Use an in-line Oo ... Oc instead, ok?
diff /usr/src
commit - 5e332dd6b4d6aadab29719a5f3abe
On Thu, 2023-03-30 at 16:31 +0200, Omar Polo wrote:
>
> Furthermore, the man page explicitly states that script -c runs
> sh(1):
>
> -c command
> Run sh -c command, instead of an interactive shell.
> [...]
>
> So I'd say that script(1) is working as intended. What was exactly
On Thu, Mar 30, 2023 at 04:43:32PM +0200, Claudio Jeker wrote:
> This diff adds a new set of ometric sets to show the protocol used to sync
> the repository. Possible values are "rrdp", "rsync", "https".
>
> This adds for example
> rpki_client_repository_protos{rpki_client_repository_protos="rrdp"
Hi,
this came up after a remark by tb@; it does a more thorough check of
the chunks in the delay list if malloc option F (which is included in
S).
Catches way more use-after-free's as otherwise we'll have to wait
until a chunk is actively re-used. With the new bucket code that might
take longer (
This diff adds a new set of ometric sets to show the protocol used to sync
the repository. Possible values are "rrdp", "rsync", "https".
This adds for example
rpki_client_repository_protos{rpki_client_repository_protos="rrdp",name="ripe",carepo="rsync://chloe.sobornost.net/rpki",notify="https://ch
Hello,
On 2023/03/29 22:39:52 +0800, lux wrote:
> Hi, the name of the shell subprocess is hardcoded in the `script'
> command.
>
> The test is as follows:
>
> $ echo $SHELL
> /bin/ksh
> $ script -c 'echo $0'
> Script started, output file is type
On Thu, Mar 30, 2023 at 03:18:36PM +0200, Claudio Jeker wrote:
> This diff cleans up extract_prefix(). The main point is to abort before
> writing anything.
That's good.
> PREFIX_SIZE() includes the length byte so exclude that again.
> I think overall the function is more readable now.
Yes, I ag
There's this "pattern" in mg where it first access(2) a path, then
does a buch of other things and finally opens it. It can do better.
Diff below removes the access() calls for conffile handling.
startupfile() now does an ffropen() (fopen() wrapper) instead of
access() and returns the file via pa
This diff cleans up extract_prefix(). The main point is to abort before
writing anything.
PREFIX_SIZE() includes the length byte so exclude that again.
I think overall the function is more readable now.
--
:wq Claudio
Index: util.c
On Thu, Mar 30, 2023 at 12:14:55PM +0200, Claudio Jeker wrote:
> This diff changes the internals of the pt_entry prefix objects a little
> bit. The goal is to allow flexible sized elements in the future.
> Also increase the size of the refcnt to 32bit and reorder the vpn specific
> objects so that
This diff changes the internals of the pt_entry prefix objects a little
bit. The goal is to allow flexible sized elements in the future.
Also increase the size of the refcnt to 32bit and reorder the vpn specific
objects so that they are more in line with each other.
--
:wq Claudio
Index: rde.h
=
20 matches
Mail list logo
