public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r13-4495] middle-end/40635 - SSA update losing PHI arg loations
@ 2022-12-05 15:51 Richard Biener
0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2022-12-05 15:51 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:0d14720f93a8139a7f234b2762c361e8e5da99cc
commit r13-4495-g0d14720f93a8139a7f234b2762c361e8e5da99cc
Author: Richard Biener <rguenther@suse.de>
Date: Mon Dec 5 16:03:21 2022 +0100
middle-end/40635 - SSA update losing PHI arg loations
The following fixes an issue where SSA update loses PHI argument
locations when updating PHI nodes it didn't create as part of the
SSA update. For the case where the reaching def is the same as
the current argument opt to do nothing and for the case where the
PHI argument already has a location keep that (that's an indication
the PHI node wasn't created as part of the update SSA process).
PR middle-end/40635
* tree-into-ssa.cc (rewrite_update_phi_arguments): Only
update the argument when the reaching definition is different
from the current argument. Keep an existing argument
location.
* gcc.dg/uninit-pr40635.c: New testcase.
Diff:
---
gcc/testsuite/gcc.dg/uninit-pr40635.c | 33 +++++++++++++++++++++++++++++++++
gcc/tree-into-ssa.cc | 11 +++++++----
2 files changed, 40 insertions(+), 4 deletions(-)
diff --git a/gcc/testsuite/gcc.dg/uninit-pr40635.c b/gcc/testsuite/gcc.dg/uninit-pr40635.c
new file mode 100644
index 00000000000..fab7c3d49d9
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/uninit-pr40635.c
@@ -0,0 +1,33 @@
+/* { dg-do compile } */
+/* { dg-options "-O -Wuninitialized" } */
+
+struct hostent {
+ char **h_addr_list;
+};
+struct hostent *gethostbyname(const char*);
+int socket(void);
+int close(int);
+int connect(int, const char*);
+
+int get_tcp_socket(const char *machine)
+{
+ struct hostent *hp;
+ int s42, x;
+ char **addr;
+
+ hp = gethostbyname(machine);
+ x = 0;
+ for (addr = hp->h_addr_list; *addr; addr++)
+ {
+ s42 = socket();
+ if (s42 < 0)
+ return -1;
+ x = connect(s42, *addr);
+ if (x == 0)
+ break;
+ close(s42);
+ }
+ if (x < 0)
+ return -1;
+ return s42; /* { dg-warning "uninitialized" } */
+}
diff --git a/gcc/tree-into-ssa.cc b/gcc/tree-into-ssa.cc
index f21ed2bea3f..9a2417d2b0b 100644
--- a/gcc/tree-into-ssa.cc
+++ b/gcc/tree-into-ssa.cc
@@ -2110,7 +2110,6 @@ rewrite_update_phi_arguments (basic_block bb)
symbol we may find NULL arguments. That's why we
take the symbol from the LHS of the PHI node. */
reaching_def = get_reaching_def (lhs_sym);
-
}
else
{
@@ -2122,8 +2121,9 @@ rewrite_update_phi_arguments (basic_block bb)
reaching_def = get_reaching_def (arg);
}
- /* Update the argument if there is a reaching def. */
- if (reaching_def)
+ /* Update the argument if there is a reaching def different
+ from arg. */
+ if (reaching_def && reaching_def != arg)
{
location_t locus;
int arg_i = PHI_ARG_INDEX_FROM_USE (arg_p);
@@ -2133,6 +2133,10 @@ rewrite_update_phi_arguments (basic_block bb)
/* Virtual operands do not need a location. */
if (virtual_operand_p (reaching_def))
locus = UNKNOWN_LOCATION;
+ /* If SSA update didn't insert this PHI the argument
+ might have a location already, keep that. */
+ else if (gimple_phi_arg_has_location (phi, arg_i))
+ locus = gimple_phi_arg_location (phi, arg_i);
else
{
gimple *stmt = SSA_NAME_DEF_STMT (reaching_def);
@@ -2150,7 +2154,6 @@ rewrite_update_phi_arguments (basic_block bb)
gimple_phi_arg_set_location (phi, arg_i, locus);
}
-
if (e->flags & EDGE_ABNORMAL)
SSA_NAME_OCCURS_IN_ABNORMAL_PHI (USE_FROM_PTR (arg_p)) = 1;
}
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-12-05 15:51 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-12-05 15:51 [gcc r13-4495] middle-end/40635 - SSA update losing PHI arg loations Richard Biener
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).