On 28/12/2022 13:49, наб wrote:
Control: retitle -1 coreutils: wc: -c st_size optimisation broken for files
larger than max(size_t)
Hi!
On Wed, Dec 28, 2022 at 01:31:12PM +0000, Pádraig Brady wrote:
On 27/12/2022 19:38, наб wrote:
To repro this, run truncate -s 2E a, then wc -c a;
this completes instantly on bullseye and spins in a read(2) loop on sid.
I think I see the issue but only on 32 bit size_t.
Ah this is x32, as per your report.
The patch doesn't apply to the Debian source unfortunately, but
I can confirm this behaves as you'd expect for a [iu]64->u32 truncation:
-- >8 --
$ truncate -s $(echo 2^32-1 | bc) a
$ time wc -c a
4294967295 a
real 0m0.002s
user 0m0.002s
sys 0m0.000s
$ truncate -s $(echo 2^32 | bc) a
$ time wc -c a
4294967296 a
real 0m0.718s
user 0m0.092s
sys 0m0.626s
-- >8 --
And I can repro this exact behaviour on i386.
Proposed upstream patch attached.
thanks,
Pádraig
From be259a269bee563e4670c5936efb9031835525ef Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C3=A1draig=20Brady?= <p...@draigbrady.com>
Date: Wed, 28 Dec 2022 14:04:19 +0000
Subject: [PATCH] wc: fix regression determining file size
* src/wc.c (wc): Use off_t rather than size_t
when calculating where to seek to, so that
we don't seek to a too low offset on systems
where size_t < off_t, which would result in
many read() calls to determine the file size.
* tests/misc/wc-proc.sh: Add a test case
sufficient for 32 bit systems at least.
Reported at https://bugs.debian.org/1027101
---
src/wc.c | 2 +-
tests/misc/wc-proc.sh | 7 +++++++
2 files changed, 8 insertions(+), 1 deletion(-)
diff --git a/src/wc.c b/src/wc.c
index bc52a8c0e..df9770396 100644
--- a/src/wc.c
+++ b/src/wc.c
@@ -431,7 +431,7 @@ wc (int fd, char const *file_x, struct fstatus *fstatus, off_t current_pos)
if (! fstatus->failed && usable_st_size (&fstatus->st)
&& 0 <= fstatus->st.st_size)
{
- size_t end_pos = fstatus->st.st_size;
+ off_t end_pos = fstatus->st.st_size;
if (current_pos < 0)
current_pos = lseek (fd, 0, SEEK_CUR);
diff --git a/tests/misc/wc-proc.sh b/tests/misc/wc-proc.sh
index 581890ddd..030872a91 100755
--- a/tests/misc/wc-proc.sh
+++ b/tests/misc/wc-proc.sh
@@ -42,4 +42,11 @@ cat <<\EOF > exp
EOF
compare exp out || fail=1
+# Ensure we don't read too much when reading,
+# as was the case on 32 bit systems
+# from coreutils-8.24 to coreutils-9.1
+if timeout 10 truncate -s1T do_read; then
+ timeout 10 wc -c do_read || fail=1
+fi
+
Exit $fail
--
2.26.2
