| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Removed duplicate #defines of AMQP_PUBLIC/AMQP_PRIVATE in amqp.h &
amqp_private.h
Split AMQP_PUBLIC into AMQP_PUBLIC_FUNCTION and AMQP_PUBLIC_VARIABLE
Added AMQP_CALL to specify calling convention (__cdecl by default)
Added -DAMQP_BUILD when building the library
Added -DAMQP_STATIC when building the library statically
(this is still incomplete on Win32, a installable amqp_config.h is
required)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Enable the following GCC options:
1. -Wstrict-prototypes
2. -Wcast-align
3. -fno-common
4. -fvisibility=hidden
This commit also includes some general cleanup of header files (mostly
for readability).
Signed-off-by: Michael Steinert <mike.steinert@gmail.com>
|
|
|
|
| |
Signed-off-by: Michael Steinert <mike.steinert@gmail.com>
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This uses memcpy to avoid the unaligned accesses. gcc on x86 seems to
do a reasonable job of optimizing them away. gcc on ARM leaves some
memcpy calls in the generated code.
It might be better in terms of the code generated to use an approach
based on packaged structs (that's wwhat the Linux kernel does). But
that would depend on compiler-specific syntax.
|
|
|
|
| |
And generally bring debian/copyright up to date.
|
| |
|
|
|
|
|
|
|
| |
(As long as you have a stdint.h and inttypes.h available.)
And kill the unclean macros in amqp.h that were only used from
test_tables.c.
|
|
|
|
|
|
|
|
| |
And fix up the resulting warnings
We don't use "-ansi -pedantic" in the tools dir, because that code
relies on libpopt, and so is unlikely ever to work with the Microsoft
compiler anyway.
|
| |
|
|
|
|
| |
They contained gcc-isms.
|
| |
|
|
|
|
| |
It relied on gccisms. Replace it with a amqp_abort function.
|
|
|
|
| |
These replace the macros that relied on gccisms.
|
| |
|
|
|
|
|
|
|
|
| |
Functions returning a heap-allocated string should return a "char *",
not a "const char *": Because the result is heap-allocated and becomes
the responsibility of the caller, it is certainly modifiable. And the
pointer will likely get passed to free(), triggering a conversion
warning from gcc. So remove all the relevant consts.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Windows doesn't generally use POSIX error codes, which poses a problem
for librabbitmq's approach of using those error codes in its API. So
make the librabbitmq error codes opaque: They are still be integers,
but client code is not supposed to assume anything about them, except
that they can be passed to a new amqp_error_string() function which
returns the corresponding error message
Internally, the error codes are either taken from a set of
librabbitmq-specific values, or correspond to an OS-specific (POSIX or
win32) error code, with a simple encoding to indicate which is which.
|
| |
|
|
|
|
|
|
| |
and accessible only through amqp_get_rpc_reply(). Add a better comment
on amqp_get_rpc_reply() to be a little clearer about when it's
inappropriate to use it.
|
| |
|
| |
|
|
|
|
|
|
|
| |
- amqp_login() no longer does amqp_channel_open() for you
- amqp_login() takes a channel_max argument
- amqp_login() actually calls amqp_tune_connection() now
- amqp_channel_close() and amqp_basic_publish() now take a channel number
|
| |
|
| |
|
| |
|
| |
|
|
|