more general build_lib script

[calad0i]

Description

$OSTYPE is not a reliable source for determining the platform. On OpenSUSE 20250503, bash returns "linux" instead of "linux-gnu" and the script failed without -fPIC.

The updated script checks if -fno-gnu-unique is supported, and appends it if so.

Type of change

• Bug fix (non-breaking change that fixes an issue)

Tests

Triggers on specific platforms.

[vloncar]

Instead of replacing "$OSTYPE" == "linux-gnu" with "$OSTYPE" == "linux"* (a few characters change) the suggestion is to invoke the compiler to do nothing just to see if it will fail and if it doesn't then add a cflag... bit of an overkill, no? 😄

[jmitrevs]

As an aside, should we move up from C++11? Vitis HLS I think uses C++14, Intel uses C++17. Probably makes little difference, though.

[vloncar]

@jmitrevs Only if we split the build script over vivado/vitis (vivado compiler will say c++14 doesn't exist)

[calad0i]

Using this way as $OSTYPE seems to unstable depending on the shell versison, and test if the compiler handles -fno-gnu-unique would always work as long as a compiler exists (and doesn't really take time).

[jmitrevs]

In general I think this fix is good and would be in favor of merging it. I just put please test as a sanity check to make sure it doesn't break things.

[jmitrevs]

What did we decide on this? I think the pytest failures are unrelated.