fix[sda]: maybe fix breaking build on gcc@11

There's a build error when building on gcc@11.3 (JSC compiler, reported by @andreasmueller):

  >> 740    tools/DT-Grid/tools/UHBDgrd.hpp:295:76: error: taking address of rvalue [-fpermissive]
     741      295 |         slice_n = static_cast<std::ostringstream*>( &(std::ostringstream() << k) )->str();

As gcc@11 switched from -std=gnu++14 to -std=gnu++17, this might fix this problem (actually the code needs to be fixed, but idc for now).

@andreasmueller will fix the offending code lines (grep shows 4 hits, but 2 are commented out, the other 2 are duplicates) to this:

     std::string out_name;
     out_name = "exclusion" + std::to_string(k) + "slice.txt";

→ hacking the build isn't really useful here.

added 1 commit

• 2717f346 - Added patch for problematic code @gcc11.3

Compare with previous version

Minimal example:

  1 #include <iostream>
  2 #include <sstream>
  3 #include <string>
  4 
  5 int main() {
  6     int k = 5;
  7 
  8     std::string slice_n, out_name;
  9     slice_n = static_cast<std::ostringstream*>( &(std::ostringstream() << k) )->str();
 10     out_name = "exclusion" + slice_n + "slice.txt";
 11 
 12     std::cout << out_name << std::endl;
 13 }
 14 

Result of gcc test.cxx:

test.cxx: In function ‘int main()’:
test.cxx:9:72: error: taking address of rvalue [-fpermissive]
    9 |     slice_n = static_cast<std::ostringstream*>( &(std::ostringstream() << k) )->str();
      |                                                  ~~~~~~~~~~~~~~~~~~~~~~^~~~~

Seems like the gcc 11 no longer allows this (independent of the standard used), although it has never been allowed by the standard itself. Nevertheless, this whole weird construct can be replaced by std::to_string(k), which is what we're doing in the patch.

@elmath: With this patch, I successfully installed sda on JURECA-DC.

merged

mentioned in commit 7c08e0fa