From f86abd2713289900f4349bc3c497ab6e9cee3c17 Mon Sep 17 00:00:00 2001 From: Mattias Andrée Date: Wed, 2 Sep 2015 02:31:05 +0200 Subject: m doc fix + info: communication utilties MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Mattias Andrée --- doc/info/mds.texinfo | 315 ++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 314 insertions(+), 1 deletion(-) (limited to 'doc/info/mds.texinfo') diff --git a/doc/info/mds.texinfo b/doc/info/mds.texinfo index 0c36968..0b60714 100644 --- a/doc/info/mds.texinfo +++ b/doc/info/mds.texinfo @@ -8707,8 +8707,13 @@ macros, but function-like macros and lvalue-macros use shorter. Inclusion-guards are formatted @code{MDS_LIBMDSCLIENT_*_H}. +libmdsclient is not yet capable of state marshalling +which is required for online updating to newer versions +of the client. + @menu * Protocol Utilties:: Low-level functions for implementing protocols. +* Communication Utilties:: Low-level communication functions. @end menu @@ -9009,7 +9014,7 @@ in the call to the function @code{libmds_next_message_id}. @code{test} is @code{NULL}. Upon successful completion zero is returned. On error -@code{-1} is returned and @code{errno} is set the indicate +@code{-1} is returned and @code{errno} is set to indicate the error. This function can fail with errno set to any error thrown @@ -9032,6 +9037,314 @@ that uses @code{va_list} instead. These functions are suffixed +@node Communication Utilties +@section Communication Utilties + +@cpindex Communication foundation +The header file @file{} provides +functions for connecting to the display server and +performing thread-safe communication. + +@cpindex Descriptor, connection +@cpindex Connection descriptor +@tpindex @code{libmds_connection_t} +@tpindex @code{struct libmds_connection} +These functions use the structured @code{libmds_connection_t}. +@{also known as @code{struct libmds_connection}@}. Instances +of this structure are referred to as connection descriptors. +This structure should be considered opaque. + +For these functions, the @code{this} parameter has the +type @code{libmds_connection_initialise* restrict}. + +@table @asis +@item @code{libmds_connection_initialise} [(@code{this}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_initialise} +Initialises a connection descriptor, but does not connect +it to the display server. + +Upon successful completion, zero is returned. On error, +@code{-1} is returned and @code{errno} is set to indicate +the error. + +This function may fail with @code{errno} set to +@code{EAGAIN}, @code{ENOMEM} or @code{EPERM}, as +specified for @code{pthread_mutex_init(3)}. + +@item @code{libmds_connection_create} [(@code{void}) @arrow{} @code{libmds_connection_t*}] +@fnindex @code{libmds_connection_create} +Wrapper for @code{libmds_connection_initialise} that +also allocates the connection descriptor. + +Upon successful completion, the descriptor is returned. +On error, @code{NULL} is returned and @code{errno} is +set to indicate the error. + +This function may fail for any reason, with @code{errno} +is to the same value, as @code{libmds_connection_initialise}. +It may also fail with @code{errno} set to @code{ENOMEM} if +the descriptor cannot be allocated. This is, on memory +exhaustion. + +@item @code{libmds_connection_destroy} [(@code{this}) @arrow{} @code{void}] +@fnindex @code{libmds_connection_destroy} +Release all resources held by a connection descriptor. +This also means that the connection to the display +server will close. Nothing will happen if @code{this} +is @code{NULL}. + +@item @code{libmds_connection_free} [(@code{this}) @arrow{} @code{void}] +@fnindex @code{libmds_connection_free} +Wrapper for @code{libmds_connection_destroy} that also +calls @code{free} on the descriptor to deallocate the +descriptor itself. + +@item @code{libmds_connection_establish} [(@code{this, const char** restrict display}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_establish} +Establishes a connection to the display server. +It is preferred to use @code{libmds_connection_establish_address}, +over using @code{libmds_connection_establish}. + +@code{display} must either be a pointer to a string, +or a pointer to @code{NULL}. It pointer @code{display} +itself must not be @code{NULL}. If @code{display} points +to @code{NULL}, it will be update with the value of +the environment variable @env{MDS_DISPLAY}, or left +@code{NULL} if not set. + +This function will parse value @code{display} points +to (after it has been updated if it points to @code{NULL},) +and attempt to connect to the display server it addresses. + +Upon successful completetion, zero is returned. On error, +@code{-1} is returned and @code{errno} is set to indicate +the error. + +This function may fail with @code{errno} set to @code{EFAULT} +if @env{MDS_DISPLAY} is not set (if @code{display} pointed +to, and will in such case still point to, @code{NULL}, or +if the display server's address is not properly formatted, +or specifies an unsupported protocol. It may also fail with +@code{errno} set to @code{ENAMETOOLONG} if the pathname of +the display server's socket is too long. The function may +also fail with any value on @code{errno} specified for +@code{socket(2)} or @code{connect(2)}, except @code{EINTR}. + +@item @code{libmds_connection_establish_address} [(@code{this, const libmds_display_address_t* restrict address}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_establish_address} +This function is an alternative to @code{libmds_connection_establish}, +which is preferred because it is easier to find out +why it potentially failed. And it can be done reliably. + +Instead of a pointer to a string formated as @env{MDS_DISPLAY}, +or a pointer to @code{NULL}. This function takes a, by +@code{libmds_parse_display_address}, preparsed address. + +Upon successful completetion, zero is returned. On error, +@code{-1} is returned and @code{errno} is set to indicate +the error. + +This function may fail with @code{errno} set to @code{EFAULT} +if the address is not full set, this can happen even if +@code{libmds_parse_display_address} returns successfully; +it means that the display server address is not properly +formatted or specifies an unsupported protocol. This function +may also fail with any value on @code{errno} specified for +@code{socket(2)} or @code{connect(2)}, except @code{EINTR}. + +@item @code{libmds_connection_send} [(@code{this, const char* restrict message, size_t length}) @arrow{} @code{size_t}] +@fnindex @code{libmds_connection_send} +This function locks a connection descriptor and --- when +it is entered the lock --- it will send a message to the +display server over the connection, and ignores all +interrupts. + +The message is specified by the parameter @code{message}, +and this length is specified by the parameter @code{length}. +Note that this function treats @code{message} as a raw +memory segment rather than as a NUL-terminated string. + +Undefined behaviour is invoked if @code{message} is +@code{NULL} or if @code{length} is zero. + +Upon successful completeion, the number of sent bytes +are returned. On error, zero is returned and @code{errno} +is set to indicate the error. + +This function may fail with errno set to any error +specified for @code{pthread_mutex_lock(3)}, or any +of the errors @code{EACCES}, @code{EWOULDBLOCK} (but +only if the socket has been modified to be nonblocking,) +@code{EBADF}, @code{ECONNRESET} --- if connection to +the display server is lost, --- @code{EDESTADDRREQ}, +@code{EFAULT}, @code{EINVAL}, @code{ENOBUFS}, @code{ENOMEM}, +@code{ENOTCONN}, @code{ENOTSOCK} or @code{EPIPE}, as +specified for @code{send(2)}. + +@item @code{libmds_connection_send_unlocked} [(@code{this, const char* restrict message, size_t length, int continue_on_interrupt}) @arrow{} @code{size_t}] +@fnindex @code{libmds_connection_send_unlocked} +Variant of @code{libmds_connection_send} that does +not perform a lock on the connection descriptor. +This is not intended for single-threaded programs, +rather, it is intended to be used instead of +@code{libmds_connection_send} if the thread already +holds a lock on the connection descriptor. + +This function is otherwise identical to +@code{libmds_connection_send}, except it has one +additional parameter: @code{continue_on_interrupt}, +and cannot fail because of failure to call +@code{pthread_mutex_lock}. It can however also fail +with @code{errno} set to @code{EINTR} if +@code{continue_on_interrupt} is zero. + +The function will continue sending the message +if it gets interrupted by a signal only if +@code{continue_on_interrupt} is non-zero. +@end table + +@file{} also provides a few +macros for use with connection descriptors. +For these macros, the @code{this} parameter has the +type @code{libmds_connection_initialise* restrict}. + +@table @asis +@item @code{libmds_connection_lock} [(@code{this}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_lock} +Wrapper for @code{pthread_mutex_lock} that locks +@code{this}'s mutex. One difference from the wrapped +function, is that rather than returning a value for +@code{errno}, it sets @code{errno} and returns zero +on success and @code{-1} on failure.@footnote{POSIX +threading and reenterant functions has a weird +property of returning an error code rather than +setting it on @code{errno}, despite that all other +C functions sets @code{errno} (provided that they +have error codes specified in @file{}.)} + +@item @code{libmds_connection_trylock} [(@code{this}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_trylock} +Wrapper for @code{pthread_mutex_trylock} that locks +@code{this}'s mutex. One difference from the wrapped +function, is that rather than returning a value for +@code{errno}, it sets @code{errno} and returns zero +on success and @code{-1} on failure. + +@item @code{libmds_connection_timedlock} [(@code{this, const struct timespec *restrict deadline}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_timedlock} +Wrapper for @code{pthread_mutex_timedlock} that locks +@code{this}'s mutex. One difference from the wrapped +function, is that rather than returning a value for +@code{errno}, it sets @code{errno} and returns zero +on success and @code{-1} on failure. +Neither parameter may be @code{NULL}. + +@item @code{libmds_connection_unlock} [(@code{this}) @arrow{} @code{int}] +@fnindex @code{libmds_connection_unlock} +Wrapper for @code{pthread_mutex_unlock} that unlocks +@code{this}'s mutex. One difference from the wrapped +function, is that rather than returning a value for +@code{errno}, it sets @code{errno} and returns zero +on success and @code{-1} on failure. + +@item @code{LIBMDS_HEADER_CLIENT_ID} [(@code{this})] +@fnindex @code{LIBMDS_HEADER_CLIENT_ID} +@fnindex @code{libmds_compose} +Macro to be used with @code{libmds_compose}, to +add the connection's client ID to a message. + +@item @code{LIBMDS_HEADER_MESSAGE_ID} [(@code{this})] +@fnindex @code{LIBMDS_HEADER_MESSAGE_ID} +@fnindex @code{libmds_next_message_id} +Macro to be used with @code{libmds_compose}, to +add the connection's next message ID to a message. +@code{libmds_next_message_id} shall have been called +prior to using this macro. + +@item @code{LIBMDS_HEADERS_STANDARD} [(@code{this})] +@fnindex @code{LIBMDS_HEADERS_STANDARD} +Macro that can be used in place of using both +@code{LIBMDS_HEADER_CLIENT_ID} and +@code{LIBMDS_HEADER_MESSAGE_ID}. +@end table + +@file{} includes @file{}, +with defines the structure @code{libmds_display_address_t} +@{also known as @code{struct libmds_display_address}@}, +and the function @code{libmds_parse_display_address}. + +@tpindex @code{libmds_display_address_t} +@tpindex @code{struct libmds_display_address} +@code{libmds_display_address_t} holds parsing results +from @code{libmds_parse_display_address}. Its members +are: + +@table @asis +@item @code{domain} [@code{int}] +The domain (protocol family), that is the first +argument for @code{socket(2)}. This is a value +whose constant is prefixed @code{PF_}.@footnote{It +is common to use @code{AF_}-constants instead, +however this practice is not portable.} +@code{-1} if not detected. + +@item @code{type} [@code{int}] +The socket type, that is the second argument +for @code{socket(2)}. This is a value +whose constant is prefixed @code{SOCK_}. +@code{-1} if not detected. + +@item @code{protocol} [@code{int}] +The protocol, that is the third argument +for @code{socket(2)}. This is a value +whose constant is affixed @code{PROTO_}. +Zero can be used for the default protocol. +For example, for stream IP sockets, the protocol +can only be TCP, thus zero would indicate +@code{IPPROTO_TCP}. +@code{-1} if not detected. + +@item @code{address} [@code{struct sockaddr*}] +The address. @code{NULL} if not detected. +You are responsible for freeing this. + +@item @code{address_len} [@code{socklen_t}] +The allocation size of @code{.address}. +This value may be set even if @code{address} +is @code{NULL}. + +@item @code{gai_error} [@code{int}] +Code for an error that has occured when parsing +the address, whose description can be retrieved +using @code{gia_strerror}, zero if none. +@end table + +@fnindex @code{libmds_parse_display_address} +The function @code{libmds_parse_display_address} +is used to parse the a display address string: + +@table @asis +@item @code{libmds_parse_display_address} [(@code{const char* restrict display, libmds_display_address_t* restrict address}) @arrow{} @code{int}] +@code{display} should be the address of the display +server, formatted as a value for @env{MDS_DISPLAY}. + +The address of the display will be stored in @code{*address}. + +Neither parameter may be @code{NULL}. + +Upon successful completion, zero is returned. On error, +@code{-1} is returned and errno is set to indicate the +error. Parsing-failure is not considered an error, +instead data that could not retrieved from the address +string will be set to @code{-1} or @code{NULL}. + +This function may fail with @code{errno} set to +@code{ENOMEM} on memory exhaustion, or @code{ENAMETOOLONG} +if the addressed socket's filename is too long. +@end table + + + @node libmdslltk @chapter libmdslltk -- cgit v1.2.3-70-g09d2