Re: [RFC] sg3_utils: sgp_dd: work around on pthread_cancel for android

[Douglas Gilbert <dgilbert@interlog.com>]

On 2017-12-22 01:24 PM, Bart Van Assche wrote:
> On Fri, 2017-12-22 at 17:00 +0000, Bean Huo (beanhuo) wrote:
>>   case "${host}" in
>> +        *-*-android*)
>> +		AC_DEFINE_UNQUOTED(SG_ON_ANDROID, 1, [sg3_utils on android])
>> +		AC_DEFINE_UNQUOTED(SG_LIB_LINUX, 1, [sg3_utils on linux])
>> +                AC_SUBST([os_cflags], [''])
>> +                AC_SUBST([os_libs], ['']) ;;
>>           *-*-linux-gnu*)
>>   		AC_DEFINE_UNQUOTED(SG_LIB_LINUX, 1, [sg3_utils on linux])
>>                   AC_SUBST([os_cflags], [''])
>> @@ -79,6 +84,7 @@ AM_CONDITIONAL(OS_OSF, [echo $host_os | grep '^osf' > /dev/null])
>>   AM_CONDITIONAL(OS_SOLARIS, [echo $host_os | grep '^solaris' > /dev/null])
>>   AM_CONDITIONAL(OS_WIN32_MINGW, [echo $host_os | grep '^mingw' > /dev/null])
>>   AM_CONDITIONAL(OS_WIN32_CYGWIN, [echo $host_os | grep '^cygwin' > /dev/null])
>> +AM_CONDITIONAL(OS_ANDROID, [echo $host_os | grep 'android' > /dev/null])
> 
> Hello Bean,
> 
> Please consider to use AC_CHECK_FUNC([pthread_cancel]) or similar to check the
> availability of pthread_cancel(). Explicit distro checks are much harder to maintain
> than checks for individual functions.

That macro doesn't work in Linux. It doesn't work in the sense that
Linux (Ubuntu 17.10) does have pthread_cancel but needs -lpthread
in the link. So AC_CHECK_FUNC([pthread_cancel]) does not place
'#define HAVE_PTHREAD_CANCEL 1' in config.h .

This seems to work in Linux but may not work in Android:
AC_SEARCH_LIBS([pthread_cancel], [pthread], [AC_DEFINE(HAVE_PTHREAD_CANCEL, 1, 
[Found pthread_cancel])], [])

Bean, could you check?

Anyway, I like having an OS type for Android (which I have renamed from
Bean's patch to SG_LIB_ANDROID to be consistent with my other OS
naming). Overall the patch works in Linux.

Doug Gilbert