Re: alternative extension: php-sm-curl

Michael via curl-and-php <[email protected]> Tue, 28 Jul 2026 17:37:12 +0500
Newsgroups gmane.comp.web.curl.php
Message-ID <CAMVeLLKm=XRrgv_DmX+6u6xMD1XXsOJPyR5rkPvfVdJ4odEQAQ@mail.gmail.com>
> So no libcurl? The name kind of implies that you use it, but maybe you meant
> apart from that?

the build script first compiles libcurl and later links it into final
php-sm-curl.dll/.so,
same as php-sm-nghttp2 and php-sm-libressl extensions are composed with
original library. so dependency is self-contained:

php-sm-nghttp2 (library + dummy extension)
php-sm-libressl (library + dummy extension)

> > Similarly, c-ares is intentionally omitted from the design; its threaded
> > asynchronous model introduces needless complexity without tangible benefit
> > over the native system resolver in an NTS architecture.
>
> This sounds like you have misunderstood something. c-ares is used *without*
> threads, it is the other way around: if you want asynch name resolving without
> c-ares (or alternative libraries) you need to use threads. That's what libcurl
> does under the hood.

c-ares is thread-based itself, it  AFAIK reports its state passively
via some event-structure,
but it is the threading model. with cooperative asynchronicity or process model
id like to base on, it would be active only when explicit call is made
(actively).

i had a problem with current curl/extension - on termination when the
network is off and
DNS request running (via threaded sys resolver) - a big delay at
termination, though
i set (with php) the CURLOPT_QUICK_EXIT - it somehow didn't work so i commented
the await in Curl_async_thrdd_destroy() function, recompiled and it
worked without a delay.
not sure what causes it, either php binding or curl.
-- 
curl-and-php mailing list
[email protected]
https://lists.haxx.se/mailman/listinfo/curl-and-php