I have been tearing my hair out for the past 3 days, trying to figure out why my C++ code outputs a "Segmentation fault". Being new to C++ didn't help either. Nor did using pthreads help with narrowing down with the problem (If you aren't able to replicate this, try it with pthreads I guess)
But now, I believe I finally have my fingers on one of the reasons why. Here's the code:
static void* pull_one_url(void* arguments)
{
struct example_struct args = *((struct example_struct *)arguments);
CURL *curl;
curl = curl_easy_init();
curl_easy_setopt(curl, CURLOPT_URL, ("https://example.com/" + (args).appendData).c_str());
curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(curl, CURLOPT_TIMEOUT_MS, 30000L);
// Set Headers
struct curl_slist *header = NULL;
header = curl_slist_append(header, ("Authorization: Bearer " + (args).myBearerToken).c_str());
header = curl_slist_append(header, "User-Agent: MyCPPApp");
curl_easy_setopt(curl, CURLOPT_HTTPHEADER, header);
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, WriteCallback);
CURLcode res = curl_easy_perform(curl);
cout << "Request sent." << endl;
return NULL;
}
Running that code in my case will throw a "Segmentation Error" and crash the app. Using any authorization bearer token seems to do that.
My question is - how do I bypass the segmentation fault? No doubt about it, my app will encounter bad auth tokens and I do NOT want it crashing because of that, especially considering that it's multithreaded and needs to continue it's work.
This may be happening because of the way I'm passing the token down from the casted pthread arguments, but I am at a loss when it comes to finding another way to do it, so if that's the case, kindly show me the right direction.
Oh, and by the way, if you know of a way to safely and easily print the output to the console right after the request in the thread finishes (it's multithreaded, without waiting for all threads to die), then please help me with that too. I am not 100% certain the code I am using right now is safe from these dreaded segmentation faults that seem to be haunting me, so I have excluded it from my testing and this post.
Related
I'm currently building a solution for a company as an intern, and I need to use the WooCommerce REST API features in my C++ project to send data to the website.
I've so far, after 2 long painful days, managed to install the cURL library (through vcpkg) and tested the library a bit with the many examples that you can find on the internet. But for now, what I found doesn't seem to match with what the people at WooCommerce put in their documentation.
For example, in this section, they show how to create a product on the platform using cURL, but I can't understand how to translate it in cURL language inside the C++ project. Heck, the command doesn't even work when I use it in the command prompt with my parameters.
#include <curl/curl.h>
#include <string>
// cUrl declaration
CURL* curl;
CURLcode res;
std::string readBuffer;
std::string URL = "http://www.example.com";
curl_global_init(CURL_GLOBAL_ALL);
curl = curl_easy_init();
if (curl) {
curl_easy_setopt(curl, CURLOPT_URL, URL);
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, WriteCallback);
curl_easy_setopt(curl, CURLOPT_WRITEDATA, &readBuffer);
res = curl_easy_perform(curl);
// Check for errors
if (res != CURLE_OK) {
std::string error = "curl_easy_perform() failed: ";
error += curl_easy_strerror(res);
error += "\nImpossible de se connecter au site WooCommerce fourni. Veuillez verifier vos paramètres et redémarrer l'application.";
wxMessageBox(error);
}
else {
std::string success = "Connexion au domaine ";
success += URL;
success += " réussie.\nPour changer de domaine, veuillez consulter la page Paramètres.";
wxMessageBox(success);
}
}
// cleanup
curl_easy_cleanup(curl);
curl_global_cleanup();
This code works fine, I know that I have to add the company's website instead of the example, but I can't figure out where to add my client key and client secret (basically like in the example shown on the WooCommerce doc). The basic cURL commands work fine in my local command prompt, but the example doesn't event work.
I know that my request for help may be kind of basic and easy to solve but I just spent the last 2 days and a half working on this and I'm starting to lose it.
Thanks for your help, I tried to speak the best english I could, so sorry in advance for any typo, or sorry if my post doesn't live up to the presentation standards of this platform, I'm kinda new around here :D
Ok, I've figured it out, for those who pass by and may have the same problem as I had. The commands you do with cURL in the terminal and with the library are totally different :
In the command prompt, you got to enter curl -X POST https://blablablabla
In the C++ library, you have to call the curl_easy_setopt() function with parameters to specify each component of the request : CURLOPT_URL is your main domain, CURLOPT_POSTFIELDS is the data you want to POST, and there are other parameters such as CURLOPT_WRITEFUNCTION, CURLOPT_WRITE_DATA,... etc. that handles the response from the server.
For me, this example was really useful, I don't know how I could have missed it :D Thanks Jesper Juhl for the advice, it is crucial to understand how HTTP and HTTPS works to figure this out.
I'm working in C++ with libcurl writing a program to interact with some APIs and I'm stuck when it comes to adding authentication info in the headers. I am new to libcurl and APIs with a basic knowledge of C++. Authentication requires an API key and a nonce hashed with HMAC_SHA256, each of which is then placed in the headers. A very simple JSON message is then sent. I've tried searching through this site but most examples seem to be in javascript or command line, and I don't see any relevant answers in them.
When I send my POST message to the server, I get a response 402 - Invalid ApiKey. My API key is 100% correct so I suspect it's something to do with the formatting or the way I've included it in the header. The site is BlinkTrade and their documentation is here, which gives some info about the header requirements.
Code snippet below:
char* message="{\"MsgType\": \"U2\",\"BalanceReqID\": 1}";
curl_easy_setopt(curl, CURLOPT_URL, "https://api.blinktrade.com/tapi/v1/message");
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, fwrite);
curl_easy_setopt(curl, CURLOPT_WRITEDATA, pFile2);
curl_easy_setopt(curl, CURLOPT_POST, 1L);
curl_easy_setopt(curl, CURLOPT_POSTFIELDS, message);
struct curl_slist *header = NULL;
header = curl_slist_append(header, "APIKey:0000000000000000000000000000000000000000");
header = curl_slist_append(header, "Nonce:1");
header = curl_slist_append(header, "Signature:1");
header = curl_slist_append(header, "Content-Type:application/json");
transfer = curl_easy_setopt(curl, CURLOPT_HTTPHEADER, header);
cout << transfer << endl;
transfer = curl_easy_perform(curl);
cout << transfer << endl;
And I get the return code 0 for curl_easy_setopt and curl_easy_perform. I've swapped the actual characters from the API key for a load of 0's, but otherwise everything is the same in terms of formatting etc. I've not actually used a hashed signature, I'll do that after I can sort out this error. I've tried adding a space after the colon and enclosing string and/or value in double quotes but I get the same response. What am I doing wrong that means my headers aren't actually recognised by the server?
Solved: the Blinktrade server returns "Invalid APIKey" not when your API key is incorrect, but when the signature is incorrect. A rather annoying mislabelling.
I get redirected to a page with address like http://example.com#foo=bar. I want to get foo=bar part of it. The whole thing would be ok too.
I found this thing:
char * url;
curl_easy_getinfo(myHandle, CURLINFO_EFFECTIVE_URL, &url);
I don't know english well to find information myself. Every time I want to find it, I find information on getting the page into string variable.
Code:
std::string readBuffer;
curl_global_init( CURL_GLOBAL_ALL);
CURL * myHandle;
CURLcode result;
myHandle = curl_easy_init();
curl_easy_setopt(myHandle, CURLOPT_COOKIEJAR, "coo.txt");
curl_easy_setopt(myHandle, CURLOPT_COOKIEFILE, "coo.txt");
curl_easy_setopt(myHandle, CURLOPT_URL, "https://www.google.ru/#q=stack");
curl_easy_setopt(myHandle, CURLOPT_WRITEFUNCTION, WriteCallback);
curl_easy_setopt(myHandle, CURLOPT_WRITEDATA, &readBuffer);
curl_easy_setopt(myHandle, CURLOPT_FOLLOWLOCATION, 1L);
result = curl_easy_perform(myHandle);
char * ch_cur_url;
result = curl_easy_getinfo(myHandle, CURLINFO_EFFECTIVE_URL,
&ch_cur_url);
printf("%s\n", ch_cur_url);
Outputs https://www.google.ru/
When I wanted https://www.google.ru/#q=stack
cURL removes the "fragment identifier" from the URL before making a request, as per the bug reports (1, 2). See also this patch. Thus the "fragment identifier" is not available as part of the CURLINFO_EFFECTIVE_URL.
If the "fragment identifier" is returned as part of a redirect (e.g. the Location HTTP header) and you can't get it any other way, then you may use the debug modes to peek on the communications between the cURL and the servers and extract the "fragment identifier" yourself. To that end you'll need to setup either CURLOPT_DEBUGFUNCTION or CURLOPT_HEADERFUNCTION.
P.S. A bit of advise: Googling the relevant information was very easy. First thing I did was to learn the "official" name of the #foo=bar. To get it I visited Wikipedia at URL and was brought to Fragment identifier. After that, Googling with the "curl fragment" netted the relevant parts. If you're looking for something, learn it's proper name.
I'm trying to download a file needed for my application off the internet (as part of installation) so that the first time the app starts up, the needed files get downloaded. For now I'm putting them on Google Drive and making them public, then I'm going to use libcURL to download them. The problem is, I just can't get the data.
I use the following link: https://docs.google.com/uc?id=documentID&export=download and replace documentID with the id. When I try connecting to the site though, it keeps giving me a small snippet of HTML code that basically says "Moved Temporarily" and gives me a link to the new URL. When I use the new link in my program, I get no output whatsoever. However, both links work just fine in my web browser, even when I'm not signed in. So Why don't they work in my program? Am I not setting up SSL options correctly, or is Google Drive simply not meant for this kind of thing?
Here's my code:
#include <curl/curl.h>
int main()
{
curl_global_init(CURL_GLOBAL_ALL);
CURL *curl;
CURLcode res;
curl = curl_easy_init();
if(curl) {
curl_easy_setopt(curl, CURLOPT_URL, "https://docs.google.com/uc?id=documentID&export=download");
curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(curl, CURLOPT_VERBOSE, 1L);
res = curl_easy_perform(curl);
}
curl_easy_cleanup(curl);
return 0;
}
Any help would be appreciated.
You'll need to set the CURLOPT_FOLLOWLOCATION option to tell cURL to follow redirects.
I do not know if this helps directly but I have always made the call
curl_global_init(CURL_GLOBAL_ALL);
which I see you don't use. I have seen this call made here in the threaded SSL code example http://curl.haxx.se/libcurl/c/threaded-ssl.html. This `curl_global_init() call will perform SSL initialisation amongst other things. It is discussed in this link http://curl.haxx.se/libcurl and also in the libcurl tutorial here http://curl.haxx.se/libcurl/c/libcurl-tutorial.html
This issue/quirk/side-effect is driving me crazy. Near the bottom the code, the response code of the HTTP interaction is passed by reference into responseCode_. However it often comes out as 0 even though the site can otherwise be accessed, and returns too quickly to be a timeout...
All variables are defined, the code below is just a snippet of a C++ method in a class. Any var_ variables are instance based. It runs on several threads, but that should not be a problem. Each class that uses libcurl has its own instance on the respective threads.
Thanks in advance for any ideas or advice...
CURL *curl;
curl = curl_easy_init();
//The URL
curl_easy_setopt(curl, CURLOPT_URL, url.getURLString().c_str());
//Timeout
curl_easy_setopt(curl, CURLOPT_TIMEOUT, &timeout_);
//disable signals to use with threads
curl_easy_setopt(curl, CURLOPT_NOSIGNAL, 1);
//Redirecting
curl_easy_setopt(curl, CURLOPT_MAXREDIRS, 5);
curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1);
//Writing callback
curl_easy_setopt(curl, CURLOPT_HEADERFUNCTION, &writerh);
curl_easy_setopt(curl, CURLOPT_HEADERDATA, &head_);
//Writing callback
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, &writerb);
curl_easy_setopt(curl, CURLOPT_WRITEDATA, &body_);
//Headers
struct curl_slist *headers = NULL;
for (std::map<std::string, std::string>::iterator itr = requestHeaders_.begin(); itr != requestHeaders_.end(); itr++) {
std::stringstream header;
header << itr->first << ": " << itr->second;
headers = curl_slist_append(headers, header.str().c_str());
}
curl_easy_setopt(curl, CURLOPT_HTTPHEADER, headers);
//UA
curl_easy_setopt(curl, CURLOPT_USERAGENT, "RDFaS-Bot/1.0 (+http://www.rdfas.com/bot)");
curl_easy_perform(curl); /* ignores error */
//Response code
curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &responseCode_);
//clean headers
curl_slist_free_all(headers);
curl_easy_cleanup(curl);
Update:
curl_easy_perform was not returning CURLE_OK when the response code was 0, as the marked answer explains. However debug hooks are very useful too and an excellent suggestion
The response code is only going to be set if curl_easy_perform() returns CURLE_OK so you should check that first to make sure curl actually performed the request successfully. Are you sure the callback functions for writing the header and body are set up correctly?
Also, make sure curl_global_init(CURL_GLOBAL_ALL) is called before these easy_perform threads start.
Assuming nothing in the curl handle returned by curl_easy_init() is shared across threads, then the code looks correct.
Use the debugging hooks built into libcurl.
Seriously. libcurl is a "C" nightmare of void*s and error codes. Everything can go wrong when using libcurl. Write your libcurl debugging hooks once and don't remove them from your code. You'll need them again, and again,... and again.