Re: [patch] Fix glitches with libstdc++ pretty printers

[Tom Tromey <tromey at redhat dot com>]

Tom> Then at link time the application Makefile could also dig around for
Tom> .py scripts and concatenate them into a new one for the application.

Daniel> For example, I'm going to ship a packaged arm-none-eabi toolchain that
Daniel> includes libstdc++.a and libstdc++-gdb.py.  If we have to complicate
Daniel> the IDE builder, and tell our users to complicate their build process,
Daniel> no one's going to get the spiffy pretty printers.  Automation really
Daniel> is everything.

I'm open to suggestions.

FWIW, this mechanism is just one possible way to load pretty printers.
There is nothing to prevent any other way.  E.g., the IDE can send
over arbitrary Python code itself.  If the IDE knows what goes into
the link, then this is not even very hard.

Tom> For Fedora I think we may pull the .py files into the separate debug
Tom> RPMs automatically.  Though, I am not completely sure -- it seems to
Tom> me that perhaps users would want to pretty-print libstdc++ containers
Tom> without necessarily installing all of its debug info.

Daniel> You've just contradicted yourself in these two statements.  If we
Daniel> don't have the debug info, would we still look for .debug/*-gdb.py?

Yeah.  I am not sure which course I want to take.

Tom