Example:
```
subroutine global_sub()
integer, dimension(4) :: iarr4=(/1,2,3,4/)
integer, dimension(4) :: jarr4
equivalence(iarr4,jarr4)
call sub1
print *, iarr4
contains
subroutine sub1
iarr4=jarr4((/4:1:-1/))
end subroutine sub1
end subroutine global_sub
```
`iarr4` and `jarr4` are equivalenced via a global aggregate storage,
but the references inside `sub1` are lowered differently.
`iarr4` is accessed via the global aggregate storage, while `jarr4`
is accessed via the argument tuple. This confuses the FIR alias
analysis,
that claims that a host associated entity cannot alias with a global
(if they have different source and do not have Target/Pointer
attributes deduced by the alias analysis).
I am not convinced that there is an issue in the alias analysis yet.
I think we'd better lower the accesses uniformly, i.e. if one variable
from an equivalence is lowered via the global aggregate storage, then
any other variable from this equivalence should be lowered the same way
(even if they are used via host association).
This patch makes sure that all symbols from an EQUIVALENCE get
and implicit SAVE attribute, if they do not have it already and
any symbol from the EQUIVALENCE is SAVEd (explicitly or implicitly).
This makes the further lowering consistent.
With this change, the following invocations will be treated as errors
(multiple actions are specified):
```
$ flang-new -fc1 -E -fsyntax-only file.95
$ flang-new -fc1 -fsyntax-only -fdebug-dump-symbols file.95
```
In the examples above it is not clear whether it is `-fsyntax-only` or
the other action that is run (i.e. `-E` or `-fdebug-dump-symbols`). It
makes sense to disallow such usage. This should also lead to cleaner and
clearer tests (the `RUN` lines using `%flang_fc1` will only allow one
action).
This change means that `flang-new -fc1` and `clang -cc1` will behave
differently when multiple action options are specified. As frontend
drivers are mostly used by compiler developers, this shouldn't affect or
confuse the compiler end-users. Also, `flang-new` and `clang` remain
consistent.
Tests are updated accordingly. More specifically, I've made sure that
every test specifies only one action. I've also taken the opportunity to
simplify "multiple-input-files.f90" a bit.
Differential Revision: https://reviews.llvm.org/D111781
The combined initializers constructed from DATA statements and explicit
static initialization in declarations needs to include derived type
component default initializations, overriding those default values
without complaint with values from explicit DATA statement or declaration
initializations when they overlap. This also has to work for objects
with storage association due to EQUIVALENCE. When storage association causes
default component initializations to overlap, emit errors if and only
if the values differ (See Fortran 2018 subclause 19.5.3, esp. paragraph
10).
The f18 front-end has a module that analyzes and converts DATA statements
into equivalent static initializers for objects. For storage-associated
objects, compiler-generated objects are created that overlay the entire
association and fill it with a combined initializer. This "data-to-inits"
module already exists, and this patch is essentially extension and
clean-up of its machinery to complete the job.
Also: emit EQUIVALENCE to module files; mark compiler-created symbols
and *don't* emit those to module files; check non-static EQUIVALENCE
sets for conflicting default component initializations, so lowering
doesn't have to check them or emit diagnostics.
Differential Revision: https://reviews.llvm.org/D109022
