Makes sense.
Thanks for the help + Rainer too

On Wed, Sep 4, 2019 at 3:45 PM Daniel Stenberg <dan...@haxx.se> wrote:

> On Wed, 4 Sep 2019, Salman Ahmed wrote:
>
> Please stop top-posting. It makes it really hard to follow this
> discussion/thread.
>
> > I updated my libcurl to 7.65.3 and at the same time I found out that the
> 3rd
> > party sdk i am using does use curl_easy_setopt(curl, CURLOPT_NOSIGNAL,
> 1).
> > After I set that to "0" my program didn't terminate when wifi was
> switched
> > off.
>
> Then you could probably achieve the same result with the older version as
> well.
>
> > But from the brief reading on google,
>
> Don't read on google. Read the genuine docs on the curl site.
>
> > it seems that CURLOPT_NOSIGNAL should be set to 1 for multithreaded
> > applications like mine. Is there a way to keep CURLOPT_NOSIGNAL = 1 and
> > handle the broken SIGPIPE problem I am getting originally?
>
> It depends on what you want to do.
>
> The SIGPIPE signal here originates from OpenSSL (when it tries to use the
> socket after the connection has closed). The only way to prevent that from
> happening in libcurl is for libcurl to switch off the signal on entry and
> switch it on again on exit. It does this if CURLOPT_NOSIGNAL is set to 1.
>
> The sighandler is however global for your application so if you use
> libcurl
> from multiple threads within that application libcurl will switch that
> state
> on/off in a racy manner and you don't really know how it'll handle the
> signal
> if/when it arrives.
>
> You're then better off adding a single global SIGPIPE ignore yourself and
> tell
> curl to hands off anything that has to do with signals.
>
> --
>
>   / daniel.haxx.se | Get the best commercial curl support there is - from
> me
>                    | Private help, bug fixes, support, ports, new features
>                    | https://www.wolfssl.com/contact/
>
-------------------------------------------------------------------
Unsubscribe: https://cool.haxx.se/list/listinfo/curl-library
Etiquette:   https://curl.haxx.se/mail/etiquette.html

Reply via email to