[Interest] QProcess::execute never returning (5.4/5.5 behaviour change)

[Frank Mertens]

On 28.07.2015 18:35, Thiago Macieira wrote:
> On Tuesday 28 July 2015 10:36:11 Alejandro Exojo wrote:
>> I probably wasn't clear enough: this is on a Mac, so I should have written
>> OSXFUSE instead of FUSE. The library is still based in Linux's, but current
>> stable OSXFUSE uses an old libfuse. And some parts might be different. This
>> is the use of signal(2) that I've found:
>>
>> https://github.com/osxfuse/fuse/blob/06018fa9d5cee5ee3e8fb7057e568f5bc8115f5
>> b/lib/mount_darwin.c#L589
>
> Oh, yeah, that's totally broken.
>
> My earlier recommendation stands then:
>
> Shoot the library in the head.
>
> (Or, if you're not inclined to, apply the sigprocmask solution from the Linux
> code. And, of course, hope that like on Linux sigprocmask works in the
> threaded case. POSIX does not guarantee it does)
>

One thing I'm wondering with all the SIGCHLD magic. How you make sure the signal handlers
do not mess around and spread EINTR in your third-party components. And I'm not just worried about
Qt's 3rd party libs.

Greetings,
FRank

-- 
Cyblogic
Web.: gplus.to/frankencode