These two patches add support for wide strings with more than INT_MAX
wide characters to vasnwprintf.
Without the change in lib/vasnprintf.c, the underlying swprintf function
is used, which always returns -1 for such arguments, and the logic in
vasnprintf.c then increases the memory allocation exponentially
(17 GB -> 34 GB -> 68 GB -> 137 GB, which finally fails).
This patch set completes step 1 of
<https://lists.gnu.org/archive/html/bug-gnulib/2024-04/msg00352.html>.
2024-06-20 Bruno Haible <br...@clisp.org>
vasnwprintf tests: Add test of huge %ls arguments.
* tests/test-vasnwprintf-big.c (main): Add tests for wide string
arguments with > 2^31 wide characters.
vasnwprintf: Support huge wide string arguments.
* lib/vasnprintf.c: (VASNPRINTF): In 64-bit builds, handle the %ls
directive ourselves.
(local_wcslen): Adjust #if condition.
>From 94d34cfb941880ba7112cefb196a4a0fc9f713de Mon Sep 17 00:00:00 2001
From: Bruno Haible <br...@clisp.org>
Date: Thu, 20 Jun 2024 08:11:53 +0200
Subject: [PATCH 1/2] vasnwprintf: Support huge wide string arguments.
* lib/vasnprintf.c: (VASNPRINTF): In 64-bit builds, handle the %ls
directive ourselves.
(local_wcslen): Adjust #if condition.
---
ChangeLog | 7 +++++++
lib/vasnprintf.c | 4 ++--
2 files changed, 9 insertions(+), 2 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index c68af8c9e7..ea01cfe60d 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,10 @@
+2024-06-20 Bruno Haible <br...@clisp.org>
+
+ vasnwprintf: Support huge wide string arguments.
+ * lib/vasnprintf.c: (VASNPRINTF): In 64-bit builds, handle the %ls
+ directive ourselves.
+ (local_wcslen): Adjust #if condition.
+
2024-06-20 Paul Eggert <egg...@cs.ucla.edu>
sigsegv: avoid unlikely undefined behavior
diff --git a/lib/vasnprintf.c b/lib/vasnprintf.c
index 1036f98498..8efb9ebee2 100644
--- a/lib/vasnprintf.c
+++ b/lib/vasnprintf.c
@@ -248,7 +248,7 @@ local_strnlen (const char *string, size_t maxlen)
# endif
#endif
-#if ((!USE_SNPRINTF || WIDE_CHAR_VERSION || !HAVE_SNPRINTF_RETVAL_C99 || USE_MSVC__SNPRINTF || NEED_WPRINTF_DIRECTIVE_LC) && WIDE_CHAR_VERSION) || ((!USE_SNPRINTF || (PTRDIFF_MAX > INT_MAX) || !HAVE_SNPRINTF_RETVAL_C99 || USE_MSVC__SNPRINTF || NEED_PRINTF_DIRECTIVE_LS) && !WIDE_CHAR_VERSION && DCHAR_IS_TCHAR)
+#if ((!USE_SNPRINTF || WIDE_CHAR_VERSION || !HAVE_SNPRINTF_RETVAL_C99 || USE_MSVC__SNPRINTF || (PTRDIFF_MAX > INT_MAX) || !DCHAR_IS_TCHAR || NEED_WPRINTF_DIRECTIVE_LC) && WIDE_CHAR_VERSION) || ((!USE_SNPRINTF || (PTRDIFF_MAX > INT_MAX) || !HAVE_SNPRINTF_RETVAL_C99 || USE_MSVC__SNPRINTF || NEED_PRINTF_DIRECTIVE_LS) && !WIDE_CHAR_VERSION && DCHAR_IS_TCHAR)
# if HAVE_WCSLEN
# define local_wcslen wcslen
# else
@@ -2993,7 +2993,7 @@ VASNPRINTF (DCHAR_T *resultbuf, size_t *lengthp,
}
}
#endif
-#if WIDE_CHAR_VERSION && (!DCHAR_IS_TCHAR || NEED_WPRINTF_DIRECTIVE_LC)
+#if WIDE_CHAR_VERSION && ((PTRDIFF_MAX > INT_MAX) || !DCHAR_IS_TCHAR || NEED_WPRINTF_DIRECTIVE_LC)
else if ((dp->conversion == 's'
&& a.arg[dp->arg_index].type == TYPE_WIDE_STRING)
|| (dp->conversion == 'c'
--
2.34.1
>From 45a1b5a72c097fc68b38638a3c56ef42c0f53d73 Mon Sep 17 00:00:00 2001
From: Bruno Haible <br...@clisp.org>
Date: Thu, 20 Jun 2024 08:13:25 +0200
Subject: [PATCH 2/2] vasnwprintf tests: Add test of huge %ls arguments.
* tests/test-vasnwprintf-big.c (main): Add tests for wide string
arguments with > 2^31 wide characters.
---
ChangeLog | 4 ++
tests/test-vasnwprintf-big.c | 91 ++++++++++++++++++++++++++++++++----
2 files changed, 87 insertions(+), 8 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index ea01cfe60d..5fea7e958b 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,9 @@
2024-06-20 Bruno Haible <br...@clisp.org>
+ vasnwprintf tests: Add test of huge %ls arguments.
+ * tests/test-vasnwprintf-big.c (main): Add tests for wide string
+ arguments with > 2^31 wide characters.
+
vasnwprintf: Support huge wide string arguments.
* lib/vasnprintf.c: (VASNPRINTF): In 64-bit builds, handle the %ls
directive ourselves.
diff --git a/tests/test-vasnwprintf-big.c b/tests/test-vasnwprintf-big.c
index 30f123f15f..1c4411efca 100644
--- a/tests/test-vasnwprintf-big.c
+++ b/tests/test-vasnwprintf-big.c
@@ -55,14 +55,16 @@ main ()
rl.rlim_cur = rl.rlim_max = 0;
setrlimit (RLIMIT_CORE, &rl);
# endif
- /* The test below needs about 25 GiB of memory:
+ /* The test below needs about 40 GiB of memory:
$ time /usr/bin/time -f "Max RSS: %M KiB" ./test-vasnwprintf-big
- Max RSS: 26216204 KiB
- real 4m22,540s
- user 4m6,322s
- sys 0m16,203s
- 5 GiB for the inputs and up to 20 GiB for temporary output buffers. */
- double needed = 25.0 * 1024 * 1024 * 1024;
+ Max RSS: 41944784 KiB
+ real 5m8,508s
+ user 4m38,035s
+ sys 0m30,456s
+ 5 GiB for the inputs in the %s tests,
+ or 20 GiB for the inputs in the %ls tests,
+ and up to 20 GiB for temporary output buffers. */
+ double needed = 40.0 * 1024 * 1024 * 1024;
double avail = physmem_claimable (1.0);
printf ("memory needed = %g MiB, available = %g MiB\n",
needed / 1024 / 1024, avail / 1024 / 1024);
@@ -70,7 +72,7 @@ main ()
{
/* Note: The malloc() calls can fail, due to ulimit of RLIMIT_DATA.
For example, on OpenBSD 7.5, the soft limit is 1.0 GiB or 1.5 GiB,
- and you need "ulimit -d 26214400". */
+ and you need "ulimit -d 42991616". */
/* Verify that asnwprintf() can return a string of size > 4 GiB. */
{
@@ -190,6 +192,79 @@ main ()
}
}
}
+
+ /* Verify that asnwprintf() can take a wide string with an element count
+ > 2^31, < 2^32 as argument. */
+ {
+ size_t n1 = 3 * (size_t) (INT_MAX / 2) + 10;
+ wchar_t *s1;
+
+ s1 = (wchar_t *) malloc ((n1 + 1) * sizeof (wchar_t));
+ if (s1 != NULL)
+ {
+ wmemset (s1, L'a', n1);
+ s1[n1] = L'\0';
+
+ size_t len;
+ wchar_t *s = asnwprintf (NULL, &len, L"x%lsy", s1);
+ if (s == NULL)
+ {
+ ASSERT (errno == ENOMEM);
+ skipped = true;
+ }
+ else
+ {
+ ASSERT (wcslen (s) == len);
+ ASSERT (len == n1 + 2);
+ size_t i;
+ for (i = 0; i <= len; i++)
+ s[i] = (i == 0 ? 'x' :
+ i <= n1 ? 'a' :
+ i == n1 + 1 ? 'y' :
+ '\0');
+ free (s);
+ }
+ free (s1);
+ }
+ }
+
+ /* Verify that asnwprintf() can take a wide string with an element count
+ > 2^32 as argument. */
+ {
+ size_t n1 = 5 * (size_t) (INT_MAX / 2) + 10;
+ if (n1 > (size_t) INT_MAX)
+ {
+ wchar_t *s1;
+
+ s1 = (wchar_t *) malloc ((n1 + 1) * sizeof (wchar_t));
+ if (s1 != NULL)
+ {
+ wmemset (s1, L'a', n1);
+ s1[n1] = L'\0';
+
+ size_t len;
+ wchar_t *s = asnwprintf (NULL, &len, L"x%lsy", s1);
+ if (s == NULL)
+ {
+ ASSERT (errno == ENOMEM);
+ skipped = true;
+ }
+ else
+ {
+ ASSERT (wcslen (s) == len);
+ ASSERT (len == n1 + 2);
+ size_t i;
+ for (i = 0; i <= len; i++)
+ s[i] = (i == 0 ? 'x' :
+ i <= n1 ? 'a' :
+ i == n1 + 1 ? 'y' :
+ '\0');
+ free (s);
+ }
+ free (s1);
+ }
+ }
+ }
}
else
skipped = true;
--
2.34.1
