HTTP server library API #140
Conversation
|
Just noticed that I forgot to fix the README.rst file in http-server, will do that in next version. |
jukkar
force-pushed
the
http-server
branch
2 times, most recently
from
6edabc0 to
67d6590
Compare
samples/net/http_server/src/main.c
Outdated
There was a problem hiding this comment.
Don't we have a k_panic() now?
samples/net/http_server/src/main.c
Outdated
There was a problem hiding this comment.
snprintf() returns an integer, and you're accumulating its return value on an unsigned short. In the minimal libc, it can return EOF, which is defined to -1; I don't know about newlib. It's better to store the return value on an integer, check for overflow (if (r < 0 || r >= size_param)), and only then accumulate to the offset variable -- making sure it's not going to go past MAX_UINT16.
subsys/net/lib/http/http_server.c
Outdated
There was a problem hiding this comment.
Doing a linear search here means that the order which URLs were added to the urls array matters when matching. You might want to consider either using a prefix tree (which is costly WRT memory), or, instead of returning the first hit, storing the last match/prefix size, and only changing that if the last stored size is larger than the current one. Alternatively you can sort them by prefix length as well and compare the larger prefixes first; just have to figure out when to sort (maybe every time a new route is added).
This way, if you had something like "/sensors" to list all the sensors in a board, and "/sensors/temperature" to get the temperature sensor, you would never match "/sensors" if asked for a specific one.
There was a problem hiding this comment.
instead of returning the first hit, storing the last match/prefix size, and only changing that if the last stored size is larger than the current one.
I'd say this is the practical solution for embedded system like Zephyr.
There was a problem hiding this comment.
This functionality comes from existing http-server application and in this in this first conversion commit (from app to lib) I would rather postpone change a bit later.
pfalcon
left a comment
pfalcon
left a comment
There was a problem hiding this comment.
So, as I found in my previous testing, this server is not compliant and doesn't allow to send HTTP headers in all cases. I imagine it expects headers to be the part of the same packet as the initial request string. Just try telnet on this server.
It also not compliant with HTTP/1.0, force-uses pipelined protocol even with it. I'd check whether Connection: close header is handled properly with HTTP/1.1 but again, I can't enter HTTP headers.
Overall, this patch is the step in the right direction, and given that it's scope is to untangle server lib from samples, I give +1, but it has a long way to go beyond that...
I fully agree, there is still much to do to get this spec compliant. The purpose of this patch was to separate the core http server functionality from sample app, no effort was made here to make fully compliant server and/or optimize the performance etc. |
The net_sample_app_init() is now able to wait that both IPv4 and IPv6 addresses are setup before continuing. Signed-off-by: Jukka Rissanen <[email protected]>
The CFLAGS and object path were set incorrectly for Bluetooth. Signed-off-by: Jukka Rissanen <[email protected]>
This helper copies desired amount of data from network packet buffer info a user provided linear buffer. Signed-off-by: Jukka Rissanen <[email protected]>
When a new UDP or TCP connection handler is to be registered, we need to check if identical handler has already been created. If a duplicate is found, the registering call will return -EALREADY. The earlier code did not check this but allowed two identical handlers to be created. The latter handler was never called in this case. Signed-off-by: Jukka Rissanen <[email protected]>
|
New version with snprintf() fixes and replacing the k_thread_spawn() with k_thread_create(). I did not change the panic() function at this point, probably the panic() could be removed all together. |
This commit creates a HTTP server library. So instead of creating a complex HTTP server application for serving HTTP requests, the developer can use the HTTP server API to create HTTP server insteances. This commit also adds support for creating HTTPS servers. Signed-off-by: Jukka Rissanen <[email protected]>
5 participants
This PR moves most of the http-server application code into a dedicated library. The library can be used to create HTTP server instances. The library supports both HTTP and HTTPS connections, and both IPv4 and IPv6 can be used at the same time. The PR modifies http-server to use this new API.