Hi!
> +Dependencies
> +------------
> +CPU partitioning is performed by the partrt tool, which is available as a git
> +submodule. Before building, do the following from the LTP root directory to
> +include rt-tools in the build:
> +
> + git submodule update utils/rt-tools
I think that you have to do git submodule init prior to this (at least
if you do this right after you cloned git repository).
> +The following build time kernel configurations must be enabled:
> +
> + CONFIG_NO_HZ_FULL=y
> + CONFIG_CPUSETS=y
> +
> +These kernel configuration parameters are also documented in the
> +READ.kernel_config file in LTP root directory.
> +
> +The following kernel boot parameter needs to be set:
> +
> + nohz_full=<cpu list>
> +
> +There might be other parameters needed to actually get nohz_full working, but
> +this is largely dependent on architecture and kernel version.
> +
> +How to run
> +----------
> +
> +The partrt_nohz_full test is run by runltp like so:
> +
> + ./runltp -f partrt_nohz_full
> +
> +To make life interesting, you can play with "-n" and "-i" options to runltp,
> to
> +generate some load in the non-realtime partition:
> +
> + ./runltp -n -i 5 -f partrt_nohz_full
> +
> +To get some more help with the cause of a failing test, do this as root
> before
> +running the test:
> +
> + echo 1 > /sys/kernel/debug/tracing/events/enable
> +
> +and then look at trace after test failed:
> +
> + cat /sys/kernel/debug/tracing/trace
> +
> +Since "count_ticks" also uses ftrace and will update tracing_cpumask to only
> +trace the nohz_full CPUs, the trace should be close to empty if the run was
> +successful.
> diff --git a/testcases/kernel/partrt/nohz_full/.gitignore
> b/testcases/kernel/partrt/nohz_full/.gitignore
> new file mode 100644
> index 0000000..5848b21
> --- /dev/null
> +++ b/testcases/kernel/partrt/nohz_full/.gitignore
> @@ -0,0 +1 @@
> +/test_partrt_nohz_full
> diff --git a/testcases/kernel/partrt/nohz_full/Makefile
> b/testcases/kernel/partrt/nohz_full/Makefile
> new file mode 100644
> index 0000000..5528dc0
> --- /dev/null
> +++ b/testcases/kernel/partrt/nohz_full/Makefile
> @@ -0,0 +1,23 @@
> +#
> +# Copyright (C) 2014, Enea AB.
> +#
> +# This program is free software; you can redistribute it and/or modify
> +# it under the terms of the GNU General Public License as published by
> +# the Free Software Foundation; either version 2 of the License, or
> +# (at your option) any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
> +# the GNU General Public License for more details.
> +#
> +
> +top_srcdir ?= ../../../..
> +
> +include $(top_srcdir)/include/mk/testcases.mk
> +
> +LDLIBS += -lrt
> +
> +MAKE_TARGETS := test_partrt_nohz_full
> +
> +include $(top_srcdir)/include/mk/generic_leaf_target.mk
> diff --git a/testcases/kernel/partrt/nohz_full/test_partrt_nohz_full.c
> b/testcases/kernel/partrt/nohz_full/test_partrt_nohz_full.c
> new file mode 100644
> index 0000000..1343b71
> --- /dev/null
> +++ b/testcases/kernel/partrt/nohz_full/test_partrt_nohz_full.c
> @@ -0,0 +1,551 @@
> +/*
> + * Copyright (C) 2014, Enea AB.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
> + * the GNU General Public License for more details.
> + */
> +
> +#define _GNU_SOURCE
> +
> +#include <errno.h>
> +#include <string.h>
> +#include <getopt.h>
> +#include <sys/types.h>
> +#include <sys/wait.h>
> +#include <sys/syscall.h>
> +#include <sys/stat.h>
> +#include <sys/timerfd.h>
> +#include <stdarg.h>
> +#include <inttypes.h>
> +#include <limits.h>
> +#include <time.h>
> +#include <sched.h>
> +#include <fcntl.h>
> +
> +#include <test.h>
> +#include <usctest.h>
> +#include <safe_macros.h>
> +#include <safe_stdio.h>
> +
> +const char *TCID = "partrt_nohz_full";
> +const int TST_TOTAL = 1;
> +
> +/* Used for RT load */
> +volatile int dummy_value;
> +
> +/* When true (1), all children will terminate */
> +static volatile int time_is_up;
> +
> +/* Verbosity level chosen from program command line.
> + * 0 = No extra information from called sub-programs
> + * 1 = Request verbose information from called sub-programs
> + * 2 = Request more verbose inormation from called sub-programs.
> + */
> +static int verbose;
The verbosity level doesn't seem to be used anywhere (apart from being
set while parsing command line options).
> +/* Name of application as given in argv[0] */
> +static const char *appname;
> +
> +/* Boolean describing whether RT partition in cpuset needs to be created */
> +int need_partrt_create = 0;
> +
> +/* RT load function prototype */
> +typedef void (child_func)(void);
> +
> +/* Amount of stack for RT load threads */
> +#define STACK_SZ (64 * 1024)
> +
> +/* Default path to cgroups/cpuset */
> +#define CPUSET_ROOT "/sys/fs/cgroup/cpuset"
> +
> +/* Expected path to cpuset RT partition */
> +#define CPUSET_RT_PATH CPUSET_ROOT "/rt"
> +
> +/* Expected path to RT partition tasks list file */
> +#define CPUSET_RT_TASKS_FILE CPUSET_RT_PATH "/tasks"
> +
> +/* Expected path to RT partition CPU list file */
> +#define CPUSET_RT_CPUS_FILE CPUSET_RT_PATH "/cpuset.cpus"
> +
> +/* File to look for to determine whether true 0Hz can be achieved */
> +#define SCHED_TICK_MAX_FILE "/sys/kernel/debug/sched_tick_max_deferment"
> +
> +/* File containing a list of filesystems supported by the kernel */
> +#define PROC_FILESYSTEMS_FILE "/proc/filesystems"
> +
> +static void cleanup(void);
> +
> +/*
> + * Check exit status value.
> + * If any signs of error is found, report error and abort testing.
> + */
> +static void assert_exit_status(const char *cmd, int status)
> +{
> + if (status == -1)
> + tst_brkm(TBROK | TERRNO, cleanup,
> + "%s: Could not execute command", cmd);
> + if (WIFSIGNALED(status))
> + tst_brkm(TBROK, cleanup,
> + "%s: Child terminated unexpectedly due to signal nr
> %d",
> + cmd,
> + WTERMSIG(status));
I think that it's more readbale to add the curly braces around block
that spawns to multiple lines even if it's technically one function
call.
> + if (!WIFEXITED(status))
> + tst_brkm(TBROK, cleanup, "%s: Child terminated unexpectedly",
> + cmd);
> + if (WEXITSTATUS(status) != 0)
> + tst_brkm(TBROK, cleanup, "%s: Child exited with exit status %d",
> + cmd,
> + WEXITSTATUS(status));
> +}
> +
> +/*
> + * Execute given command using shell, ensure success (0) is returned.
> + */
> +static void shell(const char *cmd, ...)
> +{
> + char *cmd_buf;
> + va_list va;
> +
> + va_start(va, cmd);
> + if (vasprintf(&cmd_buf, cmd, va) == -1)
> + tst_brkm(TBROK, cleanup,
> + "%s: Not valid printf format, or out of memory", cmd);
> + va_end(va);
> +
> + tst_resm(TINFO, "%s: Executing", cmd_buf);
> + assert_exit_status(cmd_buf, system(cmd_buf));
> + tst_resm(TINFO, "%s: Returns success", cmd_buf);
> +
> + free(cmd_buf);
> +}
> +
> +/*
> + * Convert string to unsigned long.
> + * Error prefix expressed as string.
> + */
> +static unsigned long str_to_ulong(
> + const char *str,
> + const char *err_prefix
> + )
Why not just:
static unsigned long str_to_ulong(const char *str, const char *err_prefix)
As far as I can the line fits into 80 chars.
> +{
> + char *endptr;
> + unsigned long val;
> + val = strtoull(str, &endptr, 0);
> + if (endptr == str)
> + tst_brkm(TBROK, cleanup,
> + "%s: %s: Expected unsigned decimal value",
> + err_prefix, str);
> + if (*endptr != '\0')
> + tst_brkm(TBROK, cleanup,
> + "%s: %s: Garbage character '%c' found after decimal
> value",
> + err_prefix, str, *endptr);
> +
> + return val;
> +}
> +
> +/*
> + * Call a command using shell.
> + * Command line expressed as va_list.
> + * Returns the command's first line of output.
> + */
> +static char *shell_str(char *dest, int size, const char *cmd, ...)
> +{
> + int len;
> + char *cmd_buf;
> + FILE *file;
> + va_list va;
> +
> + /* Build command line */
> + va_start(va, cmd);
> + if (vasprintf(&cmd_buf, cmd, va) == -1)
> + tst_brkm(TBROK, cleanup,
> + "%s: Not valid printf format, or out of memory", cmd);
> + va_end(va);
> +
> + /* Launch command */
> + errno = 0;
> + file = popen(cmd_buf, "r");
> + if (file == NULL) {
> + if (errno == 0)
> + tst_brkm(TBROK, cleanup, "%s: popen(): Out of memory",
> + cmd_buf);
> + else
> + tst_brkm(TBROK | TERRNO, cleanup,
> + "%s: popen() failed", cmd_buf);
> + }
This may be worth of SAFE_POPEN(), I can add it if you want.
> + tst_resm(TINFO, "%s: Executing", cmd_buf);
> +
> + /* Read commands stdout */
> + if (fgets(dest, size, file) == NULL) {
> + if (feof(file))
> + tst_brkm(TBROK, cleanup,
> + "%s: Expected output from the command, but got
> nothing",
> + cmd_buf);
> + else
> + tst_brkm(TBROK | TERRNO, cleanup,
> + "%s: Could not read command output",
> + cmd_buf);
> + }
> +
> + /* Get rid of terminating newline, if any */
> + len = strlen(dest);
> + if (dest[len-1] == '\n')
> + dest[len-1] = '\0';
> +
> + /* Wait until command execution finish.
> + * Main reason for doing this even though we've got what we want is for
> + * better error detection. The alternative could be to let cleanup()
> + * handle it if process hasn't finished by then. */
> + assert_exit_status(cmd_buf, pclose(file));
> +
> + tst_resm(TINFO, "%s: Returns: %s", cmd_buf, dest);
> +
> + free(cmd_buf);
> +
> + return dest;
> +}
> +
> +static int child_entry(void *arg)
> +{
> + (void) arg;
> +
> + while (!time_is_up)
> + dummy_value = rand();
> +
> + return 0;
> +}
> +
> +/*
> + * Start a new thread executing the given function, on the given cpuset
> + * partition and CPU ID. Note that the CPU ID must be one of the
> + * available CPUs in the cpuset partition.
> + */
> +static void launch_child(int cpu)
> +{
> + char *stack = SAFE_MALLOC(cleanup, STACK_SZ);
> + int tid;
> + const int cpuset_tasks_fd = SAFE_OPEN(cleanup, CPUSET_RT_TASKS_FILE,
> + O_WRONLY);
> + char *tid_str;
> + cpu_set_t cpu_set;
> +
> + tid = ltp_clone(CLONE_VM, child_entry, NULL, STACK_SZ, stack);
> + if (tid == -1)
> + tst_brkm(TBROK | TERRNO, cleanup, "clone() failed");
> +
> + /* Move child to RT partition */
> + SAFE_ASPRINTF(cleanup, &tid_str, "%ld", (long) tid);
> + SAFE_WRITE(cleanup, 1, cpuset_tasks_fd, tid_str, strlen(tid_str));
> +
> + free(tid_str);
> + SAFE_CLOSE(cleanup, cpuset_tasks_fd);
SAFE_FILE_PRINTF(cleanup, CPUSET_RT_TASKS_FILE, "%ld", (long)tid) ?
Or is there a good reason to use open(), asprintf(), write() and
close() instead?
> +
> + /* Move child to indicated CPU within the RT partition */
> + CPU_ZERO(&cpu_set);
> + CPU_SET(cpu, &cpu_set);
> + if (sched_setaffinity(tid, sizeof(cpu_set), &cpu_set) < 0)
> + tst_brkm(TBROK | TERRNO, cleanup,
> + "pid %u: sched_setaffinity() failed",
> + tid);
> +
> + tst_resm(TINFO, "pid %d: Starting RT load on cpu %d",
> + tid, cpu);
> +}
> +
> +static void cleanup(void)
> +{
> + static int cleanup_entered;
> +
> + pid_t pid;
> + int status;
> +
> + if (cleanup_entered)
> + return; /* Called from cleanup() */
> +
> + cleanup_entered = 1;
Why is this still needed? It shouldn't be if only parent process uses
the tst_* interface.
> + tst_resm(TINFO, "Cleanup: Terminating children");
> +
> + time_is_up = 1;
> +
> + do {
> + pid = wait(&status);
> + if (pid != -1) {
> + char cmd[64];
> + snprintf(cmd, sizeof(cmd), "Pid %lu",
> + (unsigned long) pid);
> + assert_exit_status(cmd, status);
> + tst_resm(TINFO,
> + "Cleanup: %s: Has terminated successfully",
> + cmd);
> + }
> + } while (pid != -1);
> +
> + if (errno != ECHILD)
> + tst_brkm(TBROK | TERRNO, NULL, "Cleanup: wait() failed");
> +
> + if (need_partrt_create)
> + shell("partrt undo");
> +
> + tst_resm(TINFO, "Cleanup: Done");
> +}
> +
> +static unsigned long determine_nohz_mask(void)
> +{
> + int range_first;
> + int range_last;
> + int bit;
> + unsigned long mask = 0;
> + FILE *stream = SAFE_FOPEN(cleanup, CPUSET_RT_CPUS_FILE, "r");
> + int nr_matches;
> +
> + for (nr_matches = fscanf(stream, "%d-%d", &range_first, &range_last);
> + nr_matches > 0;
> + nr_matches = fscanf(stream, ",%d-%d", &range_first, &range_last)) {
> + if (nr_matches == 1)
> + range_last = range_first;
> +
> + /* Set all bits in range */
> + for (bit = range_first; bit <= range_last; bit++)
> + mask |= (1 << bit);
Are you sure that this would not overflow?
I guess that it depends on how you have
partitioned your CPUs.
> + }
> +
> + SAFE_FCLOSE(cleanup, stream);
> +
> + if (nr_matches == -1)
> + tst_brkm(TBROK | TERRNO, cleanup, "%s: fscanf() failed",
> + CPUSET_RT_CPUS_FILE);
> +
> + return mask;
> +}
> +
> +static void tool_available(char *tool_name, const char *env_path)
> +{
> + int success = 0;
> +
> + if (strchr(tool_name, '/') != NULL) {
> + if (access(tool_name, X_OK) == 0)
> + success = 1;
> + } else {
> + char full_path[PATH_MAX];
> + const char *curr;
> + char * const alloced_path = strdup(env_path);
> + char *next = alloced_path;
> +
> + for (curr = strsep(&next, ":");
> + curr != NULL;
> + curr = strsep(&next, ":")) {
> + if (*curr == '\0')
> + curr = ".";
> + if (snprintf(full_path, sizeof(full_path),
> + "%s/%s", curr, tool_name
> + ) >= (int)sizeof(full_path))
> + continue;
> + if (access(full_path, X_OK) == 0) {
> + success = 1;
> + break;
> + }
> + }
> + free(alloced_path);
> + }
> +
> + if (!success)
> + tst_brkm(TCONF, cleanup,
> + "%s tool not found, skipping test. Use 'git submodule
> update utils/rt-tools' to include needed tools in the build.",
> + tool_name);
> +}
> +
> +static void tools_check(void)
> +{
> + const char *env_path = getenv("PATH");
> +
> + if (env_path == NULL)
> + env_path = "";
> +
> + tool_available("partrt", env_path);
> + tool_available("list2mask", env_path);
> + tool_available("count_ticks", env_path);
We have tst_get_path() that could be used as:
char buf[2048];
if (!tst_get_path("partrt", buf, sizeof(buf)))
tst_brkm(TCONF, cleanup, "Tool partrt not available");
> +}
> +
> +static void cpuset_check(void)
> +{
> + FILE * const stream = SAFE_FOPEN(cleanup, PROC_FILESYSTEMS_FILE, "r");
> *
> + char *name;
> +
> + while (fscanf(stream, "%as", &name) > 0) {
> + if (strcmp(name, "cpuset") == 0) {
> + free(name);
> + return;
> + }
> + free(name);
> + }
> +
> + tst_brkm(TCONF, cleanup, "CPUSET not configured in kernel");
> +}
> +
> +/*
> + * Setup and return number of child threads
> + */
> +static int setup_children(void)
> +{
> + unsigned long nohz_mask;
> + int bit;
> + int nr_children = 0;
> +
> + time_is_up = 0;
> +
> + tools_check();
> + tst_require_root(NULL);
> + cpuset_check();
> +
> + if (access(CPUSET_RT_PATH, F_OK) == -1)
> + need_partrt_create = 1;
> +
> + if (need_partrt_create)
> + shell("partrt create $(list2mask --nohz)");
> +
> + nohz_mask = determine_nohz_mask();
> + tst_resm(TINFO, "Nohz CPU mask: %#lx", nohz_mask);
> +
> + for (bit = 0; bit < (int) (sizeof(nohz_mask) * 8); bit++) {
> + if (nohz_mask & (1lu << bit)) {
> + launch_child(bit);
> + nr_children++;
> + }
> + }
> +
> + tst_resm(TINFO, "All children started");
> + return nr_children;
> +}
> +
> +/*
> + * Perform tick measurement for the given number of seconds.
> + */
> +static void test(time_t duration, unsigned nr_children)
> +{
> + uint64_t nr_ticks;
> + time_t time_finished;
> + /* If sched_tick_max_deferment patch has been applied, expect that the
> + * partitioning has disabled ticks completely. Otherwise, expect
> + * 1Hz ticks */
> + const int expect_0hz =
> + access(SCHED_TICK_MAX_FILE, F_OK) == 0;
> + const time_t current_time = time(NULL);
> +
> + const uint64_t expected_ticks = expect_0hz ? 0 : duration * nr_children;
> + char val_str[64];
> + static const char count_ticks_end_cmd[] =
> + "count_ticks --cpu rt --batch --end";
> + int timer_fd;
> + struct itimerspec timeout = { {0}, {0} };
> + uint64_t nr_timeouts;
> +
> + shell("count_ticks --cpu rt --start");
> +
> + time_finished = time(NULL) + duration;
> +
> + tst_resm(TINFO, "Execution started: %s", ctime(¤t_time));
> + tst_resm(TINFO, "Execution ends : %s", ctime(&time_finished));
What is this information good for? As far as I can see it says how long
it took to prepare for the actual test.
> + timer_fd = timerfd_create(CLOCK_MONOTONIC, 0);
> + timeout.it_value.tv_sec = duration;
> + timerfd_settime(timer_fd, 0, &timeout, NULL);
> +
> + /* Wait for timeout */
> + SAFE_READ(cleanup, 1, timer_fd, &nr_timeouts, sizeof(nr_timeouts));
> + if (nr_timeouts != 1)
> + tst_brkm(TBROK, cleanup,
> + "Multiple timeouts when single timeout was requested");
> +
> + shell_str(val_str, sizeof(val_str), count_ticks_end_cmd);
> + nr_ticks = str_to_ulong(val_str, count_ticks_end_cmd);
> +
> + if (expect_0hz) {
> + if (nr_ticks != 0)
> + tst_resm(TFAIL, "Expected no ticks, but got %" PRIu64,
> + nr_ticks);
> + else
> + tst_resm(TPASS, "No ticks occurred");
> + } else {
> + if (nr_ticks > expected_ticks)
> + tst_resm(TFAIL,
> + "Expected maximum %" PRIu64
> + " ticks, but got %" PRIu64,
> + expected_ticks, nr_ticks);
> + else
> + tst_resm(TPASS,
> + "%" PRIu64
> + " ticks occurred, which was expected",
> + nr_ticks);
> + }
> +}
> +
> +static void usage(void)
> +{
> + printf(
> + "Usage: %s [options]\n"
> + " %s --help\n"
> + "Test whether a CPU can be isolated and made tickless even
> under load.\n"
> + "\n"
> + " -h Display this usage text and exit.\n"
> + " -v <level> Set verbose level, 0 = default.\n"
> + " -d <secs> Number of seconds to run the test.\n"
> + ,
> + appname, appname
> + );
> +
> + exit(1);
The library exits after usage is printed, the exit(1) here will not be
reached.
> +}
> +
> +int main(int argc, char *argv[])
> +{
> + long duration = 0;
> + int lc;
> + unsigned int nr_children;
> + char *error_msg;
> + char *verbose_str = NULL;
> + char *duration_str = NULL;
> +
> + const option_t options[] = {
> + {"v:", NULL, &verbose_str},
> + {"d:", NULL, &duration_str},
> + {NULL, NULL, NULL}
> + };
> +
> + verbose = 0;
> + appname = basename(argv[0]);
> +
> + error_msg = parse_opts(argc, argv, options, usage);
> + if (error_msg != NULL)
> + tst_brkm(TBROK, NULL, "Error parsing command line: %s",
> + error_msg);
> +
> + if (verbose_str != NULL)
> + verbose = str_to_ulong(verbose_str, "-v");
> +
> + if (duration_str == NULL)
> + tst_brkm(TBROK, cleanup,
> + "No duration specified, nothing to do");
> +
> + duration = (long) str_to_ulong(duration_str, "-d");
> +
> + tst_resm(TINFO, "%s: Compiled %s %s", __FILE__, __DATE__, __TIME__);
I would say that this is not very useful information, or is it?
> + nr_children = setup_children();
> +
> + for (lc = 0; TEST_LOOPING(lc); lc++)
> + test(duration, nr_children);
> +
> + cleanup();
> + tst_exit();
> +
> + /* Should not end up here */
> + return 1;
Remove the return from here. The tst_exit() is marked as
__attribute__((noreturn)) and the compiler knows that you cannot get
here.
> +}
> diff --git a/utils/.gitignore b/utils/.gitignore
> index a582748..f63e51a 100644
> --- a/utils/.gitignore
> +++ b/utils/.gitignore
> @@ -48,3 +48,6 @@
> /sctp/func_tests/test_tcp_style_v6
> /sctp/func_tests/test_timetolive
> /sctp/func_tests/test_timetolive_v6
> +/count_ticks
> +/list2mask
> +/partrt
> diff --git a/utils/Makefile b/utils/Makefile
> index 1508b35..8c2fc7e 100644
> --- a/utils/Makefile
> +++ b/utils/Makefile
> @@ -22,7 +22,11 @@ top_srcdir ?= ..
>
> include $(top_srcdir)/include/mk/env_pre.mk
>
> +ifneq ($(wildcard rt-tools),)
> +MAKE_TARGETS += ffsb partrt list2mask count_ticks
> +else
> MAKE_TARGETS += ffsb
> +endif
This would be better as:
MAKE_TARGETS += ffsb
ifneq ($(wildcard rt-tools),)
MAKE_TARGETS += partrt list2mask count_ticks
endif
because otherwise we would need to add any new tool into two places
which is prone to mistakes.
> FFSBDIR := ffsb-6.0-rc2
> FILTER_OUT_DIRS := $(FFSBDIR)
> @@ -35,6 +39,9 @@ $(FFSB): $(abs_srcdir)/$(FFSBDIR)
> ffsb: $(FFSB)
> cp $(FFSB) ffsb
>
> +partrt list2mask count_ticks:
> + cp rt-tools/install/bin/$@ $@
> +
> trunk-all: $(FFSB)
>
> trunk-clean:: | ffsb-clean
> diff --git a/utils/rt-tools b/utils/rt-tools
> new file mode 160000
> index 0000000..f75b334
> --- /dev/null
> +++ b/utils/rt-tools
> @@ -0,0 +1 @@
> +Subproject commit f75b334922a2243d0b757c0627c6f1c8440818c0
> --
> 1.7.10.4
>
>
> ------------------------------------------------------------------------------
> "Accelerate Dev Cycles with Automated Cross-Browser Testing - For FREE
> Instantly run your Selenium tests across 300+ browser/OS combos. Get
> unparalleled scalability from the best Selenium testing platform available.
> Simple to use. Nothing to install. Get started now for free."
> http://p.sf.net/sfu/SauceLabs
> _______________________________________________
> Ltp-list mailing list
> Ltp-list@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/ltp-list
--
Cyril Hrubis
chru...@suse.cz
------------------------------------------------------------------------------
Is your legacy SCM system holding you back? Join Perforce May 7 to find out:
• 3 signs your SCM is hindering your productivity
• Requirements for releasing software faster
• Expert tips and advice for migrating your SCM now
http://p.sf.net/sfu/perforce
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list
