on range request, return the range and use * for the size of streams. All other
aspects should r...

because fserve init is dependent on change owner, it is not done during the normal
subsystem ini...

The headers on log content request was getting leaked.

API change is really for the allocation ...

The default max-send-size is set very high now. As block sizes are scaled based on
the incoming ...

drop the LD_PRELOAD on after fork for auth command. AIS uses this to do extra processing
for adv...

The blocks scale in size now and the limiter expands based on bitrate so we do not need to
resch...

the gnu_printf attribute is slightly more useful to us instead of printf as mingw uses the
mingw...

a filename is very varied in windows, historic from DOS, a dot at the end of a filename request
...

on override, the listener will have handle to the source set up before the fh is
dropped and pos...

The data referred to by fh could of been freed at any time after the client is made active again...

Fix setresgid() (gid not uid)

This had the effect of preventing certain things like fallback override working on
a mountpoint ...

Because the 404 route dropped the client onto the generic no_file handle, the auth lookup
failed...

There could be a fault trigger in here if the remove responses looks similar to the
add response...

after a report came in about FF playback, it seems that the accept-ranges header may be
causing ...

this would cause a small repeat effect on initial playback, the initialisation is done
either in...

For platforms that support setting MSS then allow for that to be set from the default in
cases o...

allow for the avoidance of skipping when not that many listeners on and expand
the scope for res...

We also need to make sure any intro content is not present as it could look the same
as a mini r...

this led to a retry of headers that ended up doing a derefernce of NULL

this is really for fallback to file where the initial stream is not running, we still
want a new...

should close #121, was just missing the free of memory.

keep timer updated for millisecond lookup, linux is quick these days for that but generally
it i...

don't loop on block reading, this should allow for being more even on low bitrate
streams but wi...

auto block sizing is based on incoming bitrate which works well but in cases where we
have many ...

the recent intro fix up affected the timer initialisation which would force this routine
to thin...

ID3v2 tags were just filtered out at the file processing level but as ID3v2 can occur anywhere t...

the dummy block was being triggered like for iphone as it was not taking the end range
change in...

The 0 addition into the throttle is wanted for making the limiter more even but
the main issue w...

While stream handling was ok, handling of files became a problem. This seems to work
on the late...

because of previous behaviour of <redirect> introduce this tag which can take the form
of an exa...

a range request on a rate limited non-stream content could fall foul of this.

The previous behaviour was not really that good, a simple hack for redirecting dead mounts to
ot...

in cases like iphone which request ranges like 0-1 or similar small lengths, would cause
a clien...

The previous patch worked for what it was targeted for but it does not handle all cases, so
push...

most clients can handle non-chunked non-length specified transfers without issue, but HTTP 1.1
d...

with _POSIX, the mingw system uses the C99 style printf/scanf type parameters and
avoids the pos...

if listtener gets onto the source but the source terminates before even the headers
are made the...

there is a socket active test on the keep alive kick back to worker so no need
to do it here

The rate calculation worked fine on longer durations but could suffer from some inaccuracy
when ...

keep a note of the last change in samples so that we can rewind it and replay the
sample count w...

This does not fix the timestamps so could create glitch in player but should
prevent stale point...

As FLV clients use vectors for frame lookups, if that buffer goes then that invaliates
the vecto...

in cases where things are misaligned like charset mis-matches for there is a xml
parsing issue t...

The original issue still needed addressing, the fh would get destroyed if flagged with
FS_DELETE...

expire is set to 10 seconds currently but could be longer in later releases. The shutdown
will w...

minor issue, in some cases a 1.1 client (keep alive) could leave a connection open
which could s...

In theory, they could but that can be fixed up later, for now make sure the
connection gets closed.

Disable the keep alive setting for typical stream connection, as that is unlikely to
be used aga...

do what we do for flv, only adjust the buffer pos when all of it has been written, using
the chu...

minor issue, as the memory is only created at shutdown. we just prevent the
new details being qu...

chunked response is now allowed but can be disabled via _hdr=8, but it also seems that
some play...