[Bug]: --print-chpl-settings does not pick up changes from the compiler arguments

[jabraham17]

Users currently have a number of ways to select compiler configurations with Chapel environment variables. They can set the environment variables, they can write a chplconfig file, or they can change the variable with a command line flag.

For example, CHPL_GMP=none chpl ... and chpl --gmp=none ... are equivanelt.

However, when printing out the current compiler configuration using --print-chpl-settings, these are not equivalent. CHPL_GMP=none chpl --print-chpl-settings shows the current compiler configuration with CHPL_GMP: none *, while chpl --gmp=none --print-chpl-settings shows whatever the default was.

This does not impact the correctness, as --gmp=none still works as far as selecting the right runtime build. But it can be confusing that --print-chpl-settings does not match.

[bradcray]

I'm really curious what the source of this turns out to be (particularly if I'm responsible… 😬 )

[mppf]

The compiler has a map of such environment variables. However, it doesn't look like we setenv on all of them. We do setenv on a few specific ones in compiler/main/driver.cpp. I imagine we could setenv all of these within setChapelEnvs.

[jabraham17]

Calling setenv in setChapelEnvs results in all environment variables being set explicitly, which means all environment variables have the little * after them when printed, making it meaningless. However, I think the adding setenv to setEnv should solve the problem, so setenv is only called when it is actually overriden

[mppf]

do us all a favor and rename setEnv so it's not confused with setenv :)