| Age | Commit message (Collapse) | Author |
|---|
|
Run "make update-copyright" and then...
* gnulib: Update to latest with copyright year adjusted.
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Likewise.
* tests/sample-test: Adjust to use the single most recent year.
|
|
* cfg.mk (sc_prohibit_expr_unsigned): expr uses intmax_t internally
when GMP is not available, so flag any uses of unsigned limits.
* tests/misc/timeout-parameters.sh: Remove the overflow related
to UINT_MAX as it's handled by the following case. Change the
following case from 99... to $TIME_T_OFLOW to be more expressive
in what it's actually testing. Directly check that commands succeed,
rather than inspecting $? afterwards.
* tests/dd/skip-seek-past-dev.sh: Guard against large device sizes.
* tests/id/setgid.sh: Protect: Guard against large group IDs.
* tests/misc/cut-huge-range.sh: Tweak comment to avoid syntax check.
It's tempting to simplify to just skip the test if expr fails,
but SIZE_MAX > INTMAX_MAX is the common case.
|
|
Run "make update-copyright" and then...
* gnulib: Update to latest with copyright year adjusted.
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Likewise.
* tests/sample-test: Adjust to use the single most recent year.
|
|
* init.cfg (get_min_ulimit_v_): Increase the determined
virtual memory limit for the command by a page size
to avoid alignment issues causing false failures for
subsequent runs.
* tests/misc/cut-huge-range.sh: Be defensive and match
the cut invocations under test, more closely with the
form used to determine the VM memory limit.
This was noticed on i686 linux at:
http://hydra.nixos.org/build/28990456
|
|
* tests/dd/no-allocate.sh: Account for timeout(1) when
determining the required mem, as timeout has additional shared libs.
This avoids the need for the hardcoded 4M addition to the limit.
* tests/misc/head-c.sh: Increase the base limit, to account for
the fact that head(1) will allocate some additional mem in this case.
* tests/misc/cut-huge-range.sh: Remove mention of specific limits.
* tests/misc/printf-surprise.sh: Likewise.
Reported by Dmitry V. Levin
|
|
When configured with either 'symlinks' or 'shebangs' as value for
the --enable-single-binary option, tests based on `ulimit -v` are
skipped. The reason is that the multicall 'coreutils' binary requires
much more memory due to shared libraries being loaded, and the size of
the 'date' binary (~290KiB) compared to the multicall binary (~5MiB),
of course. Finally, in the case of 'shebangs', the starting shell
requires more memory, too
Instead of using hard-coded values for the memory limit, use an
adaptive approach: first determine the amount of memory for a similar,
yet more trivial invocation of the command, and then do the real test
run using that limit (plus some buffer in some cases).
* init.cfg (require_ulimit_v_): Remove function.
(get_min_ulimit_v_): Add function to determine the minimum memory limit
required for a given command in an adaptive way.
* cfg.mk (sc_prohibit_test_ulimit_without_require_): Change the name
of the above function in the syntax-check rule.
* tests/cp/link-heap.sh: Use the above function to determine the
minimum memory required to run a command simpler than in the real test
run. Use that limit plus a buffer there. While at it, change to list
of commands in the subshell to fail also if the beginning `ulimit -v`
fails.
* tests/dd/no-allocate.sh: Likewise.
* tests/misc/csplit-heap.sh: Likewise.
* tests/misc/cut-huge-range.sh: Likewise.
* tests/misc/head-c.sh: Likewise.
* tests/misc/printf-surprise.sh: Likewise.
* tests/split/line-bytes.sh: Likewise.
* tests/rm/many-dir-entries-vs-OOM.sh: Likewise - doing it separately
for each program under test.
|
|
Many tests use `program ... && fail=1` to ensure expected
error situations are indicated. However that would mask
an unexpected exit (like a crash). Therefore explicitly
check the expected exit code.
Note where error messages are also verified, the extra
protection is not added.
* tests/init.sh (returns_): A new helper function to
check the return code of a command, and used
throughout the tests.
* cfg.mk (sc_prohibit_and_fail_1): Add a syntax check
to avoid new instances of this issue.
|
|
Run "make update-copyright" and then...
* tests/sample-test: Adjust to use the single most recent year.
* tests/du/bind-mount-dir-cycle-v2.sh: Fix case in copyright message,
so that year is updated automatically in future.
|
|
Run "make update-copyright", but then also run this,
perl -pi -e 's/2\d\d\d-//' tests/sample-test
to make that one script use the single most recent year number.
|
|
* tests/misc/cut-huge-range.sh (subtract_one): New string.
(CUT_MAX): Don't pass a too-large integer to 'expr'.
|
|
Use a sentinel value that's checked implicitly, rather than
a bit array, to determine if an item should be output.
Benchmark results for this change are:
$ yes abcdfeg | head -n1MB > big-file
$ for c in orig sentinel; do
src/cut-$c 2>/dev/null
echo -ne "\n== $c =="
time src/cut-$c -b1,3 big-file > /dev/null
done
== orig ==
real 0m0.049s
user 0m0.044s
sys 0m0.005s
== sentinel ==
real 0m0.035s
user 0m0.032s
sys 0m0.002s
## Again with --output-delimiter ##
$ for c in orig sentinel; do
src/cut-$c 2>/dev/null
echo -ne "\n== $c =="
time src/cut-$c -b1,3 --output-delimiter=: big-file > /dev/null
done
== orig ==
real 0m0.106s
user 0m0.103s
sys 0m0.002s
== sentinel ==
real 0m0.055s
user 0m0.052s
sys 0m0.003s
eol_range_start: Removed. 'n-' is no longer treated specially,
and instead SIZE_MAX is set for the 'hi' limit, and tested implicitly.
complement_rp: Used to complement 'rp' when '--complement' is specified.
ADD_RANGE_PAIR: Macro renamed to 'add_range_pair' function.
* tests/misc/cut-huge-range.sh: Adjust to the SENTINEL value.
Also remove the overlapping range test as this is no longer
dependent on large ranges and also is already handled with
the EOL-subsumed-3 test in cut.pl.
|
|
This issue was introduced in commit v8.21-43-g3e466ad
* src/cut.c (set_fields): Process all range pairs when merging.
* tests/misc/cut-huge-range.sh: Add a test for this edge case.
Also fix an issue where we could miss reported errors due
to truncation of the 'err' file.
|
|
print_kth() is the central function of cut used to
determine if an item is to be output or not,
so simplify it by moving some logic outside.
Benchmark results for this change are:
$ yes abcdfeg | head -n1MB > big-file
$ for c in orig split; do
src/cut-$c 2>/dev/null
echo -ne "\n== $c =="
time src/cut-$c -b1,3 big-file > /dev/null
done
== orig ==
real 0m0.111s
user 0m0.108s
sys 0m0.002s
== split ==
real 0m0.088s
user 0m0.081s
sys 0m0.007s
* src/cut.c (print_kth): Refactor a branch to outside the function.
Related to http://bugs.gnu.org/13127
|
|
The current implementation of cut, uses a bit array,
an array of `struct range_pair's, and (when --output-delimiter
is specified) a hash_table. The new implementation will use
only an array of `struct range_pair's.
The old implementation is memory inefficient because:
1. When -b with a big num is specified, it allocates a lot of
memory for `printable_field'.
2. When --output-delimiter is specified, it will allocate 31 buckets.
Even if only a few ranges are specified.
Note CPU overhead is increased to determine if an item is to be printed,
as shown by:
$ yes abcdfeg | head -n1MB > big-file
$ for c in with-bitarray without-bitarray; do
src/cut-$c 2>/dev/null
echo -ne "\n== $c =="
time src/cut-$c -b1,3 big-file > /dev/null
done
== with-bitarray ==
real 0m0.084s
user 0m0.078s
sys 0m0.006s
== without-bitarray ==
real 0m0.111s
user 0m0.108s
sys 0m0.002s
Subsequent patches will reduce this overhead.
* src/cut.c (set_fields): Set and initialize RP
instead of printable_field.
* src/cut.c (is_range_start_index): Use CURRENT_RP rather than a hash.
* tests/misc/cut.pl: Check if `eol_range_start' is set correctly.
* tests/misc/cut-huge-range.sh: Rename from cut-huge-to-eol-range.sh,
and add a test to verify large amounts of mem aren't allocated.
Fixes http://bugs.gnu.org/13127
|
