I wonder how we deal with this issue in packages that have hand-written
Makefiles;
The issue is that some versions of gcc/g++ (if used as a linker) need
-pthread option to be specified, apart from linking
with -lpthread.
In particular this has to be dealt on https://trac.sagemath.org/ticket/2335
-pthread like -openmp should always be the right option to use. gcc/g++ and
other
are compiler drivers (the real compilers are cc1/cc1++) and those switches
tells
the driver to set everything right for compiling and linking with these
features.
That’s the theory at least. Do you have an examp
On 2018-04-20 13:48, Dima Pasechnik wrote:
The issue is that some versions of gcc/g++ (if used as a linker) need
-pthread option to be specified
That's a true statement but why do you call it an "issue"?
The GCC documentation says
-pthread
Adds support for multithreading wit
On Friday, April 20, 2018 at 1:00:56 PM UTC+1, François Bissey wrote:
>
> -pthread like -openmp should always be the right option to use. gcc/g++
> and other
> are compiler drivers (the real compilers are cc1/cc1++) and those switches
> tells
> the driver to set everything right for compili
If Sage calls Maxima, then it is a bug in Sage also.
Interestingly, Maxima gets the INdefinite integral correct.
On Sunday, April 15, 2018 at 10:31:36 PM UTC-7, Ralf Stephan wrote:
>
> Not Sage, it's Maxima:
> (%i2) integrate(sin(x)*exp(%i*x),x,-%pi,0);
>log
> On 21/04/2018, at 02:43, Dima Pasechnik wrote:
>
> clang -pthread does not do what gcc -pthread does, see
> https://stackoverflow.com/a/19382746/557937
>
> This means that with clang you need -lpthread
> whereas with gcc you need -pthread
>
> It's probably OK to have them both, no?
>
Th