apr_vformatter() is a generic printf-style formatting routine with some extensions.
- Parameters
-
flush_func | The function to call when the buffer is full |
c | The buffer to write to |
fmt | The format string |
ap | The arguments to use to fill out the format string. |
%pA takes a struct in_addr *, and prints it as a.b.c.d
%pI takes an apr_sockaddr_t * and prints it as a.b.c.d:port or
[ipv6-address]:port
%pT takes an apr_os_thread_t * and prints it in decimal
('0' is printed if !APR_HAS_THREADS)
%pt takes an apr_os_thread_t * and prints it in hexadecimal
('0' is printed if !APR_HAS_THREADS)
%pm takes an apr_status_t * and prints the appropriate error
string (from apr_strerror) corresponding to that error code.
%pp takes a void * and outputs it in hex
%pB takes a apr_uint32_t * as bytes and outputs it's apr_strfsize
%pF same as above, but takes a apr_off_t *
%pS same as above, but takes a apr_size_t *
%pA, %pI, %pT, %pp are available from APR 1.0.0 onwards (and in 0.9.x).
%pt is only available from APR 1.2.0 onwards.
%pm, %pB, %pF and %pS are only available from APR 1.3.0 onwards.
The %p hacks are to force gcc's printf warning code to skip
over a pointer argument without complaining. This does
mean that the ANSI-style %p (output a void * in hex format) won't
work as expected at all, but that seems to be a fair trade-off
for the increased robustness of having printf-warnings work.
Additionally, apr_vformatter allows for arbitrary output methods
using the apr_vformatter_buff and flush_func.
The apr_vformatter_buff has two elements curpos and endpos.
curpos is where apr_vformatter will write the next byte of output.
It proceeds writing output to curpos, and updating curpos, until
either the end of output is reached, or curpos == endpos (i.e. the
buffer is full).
If the end of output is reached, apr_vformatter returns the
number of bytes written.
When the buffer is full, the flush_func is called. The flush_func
can return -1 to indicate that no further output should be attempted,
and apr_vformatter will return immediately with -1. Otherwise
the flush_func should flush the buffer in whatever manner is
appropriate, re apr_pool_t nitialize curpos and endpos, and return 0.
Note that flush_func is only invoked as a result of attempting to
write another byte at curpos when curpos >= endpos. So for
example, it's possible when the output exactly matches the buffer
space available that curpos == endpos will be true when
apr_vformatter returns.
apr_vformatter does not call out to any other code, it is entirely
self-contained. This allows the callers to do things which are
otherwise "unsafe". For example, apr_psprintf uses the "scratch"
space at the unallocated end of a block, and doesn't actually
complete the allocation until apr_vformatter returns. apr_psprintf
would be completely broken if apr_vformatter were to call anything
that used this same pool. Similarly http_bprintf() uses the "scratch"
space at the end of its output buffer, and doesn't actually note
that the space is in use until it either has to flush the buffer
or until apr_vformatter returns.