[Freeswitch-users] Spidermokey_curl problem? Sanity check please

[Simon Tang]

Hello,

 

I've been making Javascript curl requests and have found that larger
requests sometimes come back incomplete.  This is what I am doing in
javascript:

------------

function callback(string, arg)

{

    console_log("info", "returned body:" + string + "\n");

    return true;

}

 

var curl = new CURL();

var blah = curl.run("GET", url, args, callback, arg, "");

--------------

 

comparing 'string' with tcpdump, I found that 'string' is definitely
incomplete vs. tcpdump.

 

I've looked into spidermokey_curl.c and found this code (excerpts):

 

static size_t file_callback(void *ptr, size_t size, size_t nmemb, void
*data)

{...}

 

curl_easy_setopt(co->curl_handle, CURLOPT_WRITEFUNCTION, file_callback);

curl_easy_setopt(co->curl_handle, CURLOPT_WRITEDATA, (void *) co);

 

I had the suspicion that the callback is getting a partial response and
passing it into the JS object as a completed response:

 

-          From
http://curl.netmirror.org/libcurl/c/curl_easy_setopt.html: "The callback
function will be passed as much data as possible in all invokes, but you
cannot possibly make any assumptions. It may be one byte, it may be
thousands. The maximum amount of data that can be passed to the write
callback is defined in the curl.h header file: CURL_MAX_WRITE_SIZE."
>From this, I think that it is the callback's responsibility to ensure
that the complete response was received and to save the data somewhere.

 

To confirm my suspicion, I replaced the callback function in
spidermonkey_curl.c with vanilla logging and took out the JSAPI
references:

 

static size_t file_callback(void *ptr, size_t size, size_t nmemb, void
*data)

{

    register unsigned int realsize = (unsigned int) (size * nmemb);

switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG,"BODY
STUFF::::::%s\n", (char*) ptr);

return realsize;

}

 

I recompiled and reproduced the offending http request.  I found that
the log is called twice, with each log comprising both pieces of the
complete response.  When I reverted back to the original file_callback
function and added a log in there, the log only recorded the first part
of the response (incomplete).

 

I'm not sure if I'm using this correctly or if my analysis is correct,
so if someone can provide me some feedback, that'd be great.  If it's a
legit bug, I can file it.

 

Simon Tang

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20080715/c116e65b/attachment-0002.html