Rather than storing the relations between free-regions in a global
table, introduce a `FreeRegionMap` data structure. regionck computes the
`FreeRegionMap` for each fn and stores the result into the tcx so that
borrowck can use it (this could perhaps be refactored to have borrowck
recompute the map, but it's a bid tedious to recompute due to the
interaction of closures and free fns). The main reason to do this is
because of #22779 -- using a global table was incorrect because when
validating impl method signatures, we want to use the free region
relationships from the *trait*, not the impl.
Fixes#22779.
It was in pretty good shape, but since that is my pet peeve, I clarified the compiler/interpreter distinction and why it is irrelevant for this section. Otherwise only a couple of minor clarifications, and weasel words where reality is more complicated than the text accounted for (e.g., there is more than one kind of library).
r? @steveklabnik
`ToCStr` was removed with `old_io` and the current method `as_os_str`
is inherent to `Path`, meaning there is no suitable trait bound that
could be used here.
r? @alexcrichton
1) Moved 'while' section below 'loop', 'break', and 'continue'; 2) Added information to 'while' and 'for' loops that they interact with 'break' and 'continue' and may have a lifetime label. 3) Clarified labeling syntax on the infinite loops.
(And other kinds of items).
This brings trait/impl items in line with top-level items in this respect. It means we don't get the attributes when showing error messages for methods, etc. (a good thing, I think). Since attributes have spans it is still easy to create a span for the item + all attributes.
1) Moved 'while' section below 'loop', 'break', and 'continue'; 2) Added information to 'while' and 'for' loops that they interact with 'break' and 'continue' and may have a lifetime label. 3) Clarified labeling syntax on the infinite loops.
skolemize_late_bound_regions essentially copies the entire type (most of the times it shouldn't, but it does), and match_impl runs millions of times.
Times compiling rustc, tested with
$ make -j4 rustc-stage1
$ ( time RUSTFLAGS=-Z time-passes make -j4 rustc-stage2 ) # need LLVM time for calibration
Before:
real 21m44.960s
user 29m38.812s
sys 0m14.944s
After:
real 19m31.445s
user 26m47.260s
sys 0m14.952s
Making this is a 10% performance improvement.
LLVM passes took 867 seconds before, 862 seconds after.
This required fixing the `pretty-rpass-full` tests to have the same `$$(CSREQ$(1)_T_$(2)_H_$(3))` dependencies as the `rpass-full` and `cfail-full` tests. It also required fixing the `run-make/simd-ffi` test to use unique names for its output files.
