On 2/2/2022 7:42 AM, Richard Biener via Gcc-patches wrote:
This adds a flag to CONSTRUCTOR nodes indicating that for
clobbers this marks the end-of-life of storage as opposed to
just ending the lifetime of the object that occupied it.
The dangling pointer diagnostics uses CLOBBERs but is confused
by those emitted by the C++ frontend for example which emits
them for the second purpose at the start of CTORs. The issue
is also appearant for aarch64 in PR104092.
Distinguishing the two cases is also necessary for the PR90348 fix.
Bootstrapped and tested on x86_64-unknown-linux-gnu. I verified
the bogus diagnostic in PR104092 is gone with a cross.
OK for trunk?
Thanks,
Richard.
2022-02-02 Richard Biener <rguent...@suse.de>
PR middle-end/90348
PR middle-end/104092
* tree-core.h: Document protected_flag use on CONSTRUCTOR nodes.
* tree.h (CLOBBER_MARKS_EOL): Add.
* tree-pretty-print.cc (dump_generic_node): Mark EOL CLOBBERs.
* gimplify.c (gimplify_bind_expr): Mark storage end-of-life clobbers
with CLOBBER_MARKS_EOL.
(gimplify_target_expr): Likewise.
* tree-inline.cc (expand_call_inline): Likewise.
* tree-ssa-ccp.cc (insert_clobber_before_stack_restore): Likewise.
* gimple-ssa-warn-access.cc (pass_waccess::check_stmt): Only treat
CLOBBER_MARKS_EOL clobbers as ending lifetime of storage.
* gcc.dg/pr87052.c: Adjust.
OK. Note that I think something similar may be needed to mark EOL for
the pointer passed to realloc to fix a related set of false positives
for code like this
bool something = p != q;
whatever = realloc (p, newsize)
if (something)
We forward propagate the p != q test generating something like this;
whatever - realloc (p, newsize);
if (p != q)
Which triggers a warning even though the original source was
reasonable. I think a well placed clobber of p should expose the
dataflow necessary to prevent the propagation and ultimately avoid the
false positive. IIRC something like this came up in glibc and/or gnulib.
I realize it's not exactly the same, but they're reasonably closely related.
jeff
