On 08/17/2018 03:14 AM, Bernd Edlinger wrote: > Hi! > > > After the other patch has been applied, I re-based this patch accordingly. > > Except the mechanical changes, there are a few notable differences to the > previous version: > > In string_constant, I added a similar check for the STRING_CSTs > because when callers don't use mem_size, they assume to be > able to read "TREE_STRING_LENGTH (array)" bytes, but that is > not always the case, for languages that don't always use > zero-terminated strings, for instance hollerith strings in fortran. > > --- gcc/expr.c 2018-08-17 05:32:57.332211963 +0200 > +++ gcc/expr.c 2018-08-16 23:08:23.544940795 +0200 > @@ -11372,6 +11372,9 @@ string_constant (tree arg, tree *ptr_off > *ptr_offset = fold_convert (sizetype, offset); > if (mem_size) > *mem_size = TYPE_SIZE_UNIT (TREE_TYPE (array)); > + else if (compare_tree_int (TYPE_SIZE_UNIT (TREE_TYPE (array)), > + TREE_STRING_LENGTH (array)) < 0) > + return NULL_TREE; > return array; > } > > > The range check in c_getstr was refined as well: > > This I added, because vla arrays can be initialized with string constants, > especially since the 71625 patch was installed: > In this case we end up with mem_size that fails to be constant. > > > @@ -14606,25 +14603,17 @@ c_getstr (tree src, unsigned HOST_WIDE_I > offset = tree_to_uhwi (offset_node); > } > > + if (!tree_fits_uhwi_p (mem_size)) > + return NULL; > + > /* STRING_LENGTH is the size of the string literal, including any > embedded NULs. STRING_SIZE is the size of the array the string > literal is stored in. */ > > Also the rest of the string length checks are refined, to return > actually zero-terminated single byte strings when strlen is not given, > and return something not necessarily zero-terminated which is > suitable for memxxx-functions otherwise. > > > Bootstrapped and reg-tested on x86_64-pc-linux-gnu. > Is it OK for trunk? > > > Thanks > Bernd. > > > patch-pr86711.diff > > > gcc: > 2018-08-17 Bernd Edlinger <bernd.edlin...@hotmail.de> > > PR middle-end/86711 > PR middle-end/86714 > * expr.c (string_constant): Don't return truncated string literals. > * fold-const.c (c_getstr): Fix function comment. Remove unused third > argument. Fix range checks. > * fold-const.c (c_getstr): Adjust protoype. > > testsuite: > 2018-08-17 Bernd Edlinger <bernd.edlin...@hotmail.de> > > PR middle-end/86711 > PR middle-end/86714 > * gcc.c-torture/execute/pr86711.c: New test. > * gcc.c-torture/execute/pr86714.c: New test.
Note that Martin's patch covers both these tests in slightly better ways. Jeff
