[Bug driver/87758] --print-file-name= ignores -L

[John.Ericson at Obsidian dot Systems]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87758

--- Comment #2 from John Ericson <John.Ericson at Obsidian dot Systems> ---
Indeed it has. Are you worried about the breakage then? Could we make a new
--print-library-name= ? The old name is pretty bad anyways, so I'd like a flag
with that name to exist regardless.

I am still unclear on the degree to which -L is really a GCC concern vs just
something forwarded to the linker. If the latter, maybe I should also open a
binutils issue so as not to induce a layer violation? The GCC
--print-library-name= would use the linker --print-library-name= so as to not
track something (-L path) it otherwise doesn't need to track.  

------

Also, if all this sounds like a rather theoretical layer violation,
https://ghc.haskell.org/trac/ghc/timeline?from=2018-10-26T02%3A26%3A15Z&precision=second
is an example of a practical ramification of this. The gist of this scenario is
we have another compiler, GHC, that uses gcc. Wrapping gcc to add -L is a leaky
abstraction because GHC uses --print-file-name to get at the library location.
It itself will turn its own -L into -B to get around this issue, but any
interpositioned -L will do nothing. Of course the warpper script could also use
-B, but that sure feels like a sledgehammer to solve this problem.