Bug and patch for incorrect handling of shell output with eol-strip
A full pull request with comments and detailed description of the bug
is available on the boostorg github:
https://github.com/boostorg/build/pull/136. I'm led to believe that
the preferred contribution method is via this list instead.
In summary, when the output of a shell command exceeds 1024 characters
and the strip-eol option is selected and the 1024th character (or
modulo of) is whitespace, it is trimmed due to incorrect parsing in
the `builtin_shell` function. This is especially important for shell
commands that produce compiler/linker flags such as pkg-config.
Attached is a patch with unit tests that cover the scenario and fix.
The only problem I had when writing the patch was determining if the C
unit tests run. How should I verify these?