diff options
| author | Nat Goodspeed <nat@lindenlab.com> | 2012-01-20 18:10:40 -0500 |
|---|---|---|
| committer | Nat Goodspeed <nat@lindenlab.com> | 2012-01-20 18:10:40 -0500 |
| commit | f0dbb878337082d3f581874c12e6df2f4659a464 (patch) | |
| tree | d147656500ef78b10b2fe6176b03f617a6862429 /indra/llcommon/tests/llprocess_test.cpp | |
| parent | 083a9e0927144a9e2f052bc8573da8a26259a257 (diff) | |
Per Richard, replace LLProcessLauncher with LLProcess.
LLProcessLauncher had the somewhat fuzzy mandate of (1) accumulating
parameters with which to launch a child process and (2) sometimes tracking the
lifespan of the ensuing child process. But a valid LLProcessLauncher object
might or might not have ever been associated with an actual child process.
LLProcess specifically tracks a child process. In effect, it's a fairly thin
wrapper around a process HANDLE (on Windows) or pid_t (elsewhere), with
lifespan management thrown in. A static LLProcess::create() method launches a
new child; create() accepts an LLSD bundle with child parameters. So building
up a parameter bundle is deferred to LLSD rather than conflated with the
process management object.
Reconcile all known LLProcessLauncher consumers in the viewer code base,
notably the class unit tests.
Diffstat (limited to 'indra/llcommon/tests/llprocess_test.cpp')
| -rw-r--r-- | indra/llcommon/tests/llprocess_test.cpp | 706 |
1 files changed, 706 insertions, 0 deletions
diff --git a/indra/llcommon/tests/llprocess_test.cpp b/indra/llcommon/tests/llprocess_test.cpp new file mode 100644 index 0000000000..55e22abd81 --- /dev/null +++ b/indra/llcommon/tests/llprocess_test.cpp @@ -0,0 +1,706 @@ +/** + * @file llprocess_test.cpp + * @author Nat Goodspeed + * @date 2011-12-19 + * @brief Test for llprocess. + * + * $LicenseInfo:firstyear=2011&license=viewerlgpl$ + * Copyright (c) 2011, Linden Research, Inc. + * $/LicenseInfo$ + */ + +// Precompiled header +#include "linden_common.h" +// associated header +#include "llprocess.h" +// STL headers +#include <vector> +#include <list> +// std headers +#include <fstream> +// external library headers +#include "llapr.h" +#include "apr_thread_proc.h" +#include <boost/foreach.hpp> +#include <boost/function.hpp> +#include <boost/algorithm/string/find_iterator.hpp> +#include <boost/algorithm/string/finder.hpp> +//#include <boost/lambda/lambda.hpp> +//#include <boost/lambda/bind.hpp> +// other Linden headers +#include "../test/lltut.h" +#include "../test/manageapr.h" +#include "../test/namedtempfile.h" +#include "stringize.h" +#include "llsdutil.h" + +#if defined(LL_WINDOWS) +#define sleep(secs) _sleep((secs) * 1000) +#define EOL "\r\n" +#else +#define EOL "\n" +#include <sys/wait.h> +#endif + +//namespace lambda = boost::lambda; + +// static instance of this manages APR init/cleanup +static ManageAPR manager; + +/***************************************************************************** +* Helpers +*****************************************************************************/ + +#define ensure_equals_(left, right) \ + ensure_equals(STRINGIZE(#left << " != " << #right), (left), (right)) + +#define aprchk(expr) aprchk_(#expr, (expr)) +static void aprchk_(const char* call, apr_status_t rv, apr_status_t expected=APR_SUCCESS) +{ + tut::ensure_equals(STRINGIZE(call << " => " << rv << ": " << manager.strerror(rv)), + rv, expected); +} + +/** + * Read specified file using std::getline(). It is assumed to be an error if + * the file is empty: don't use this function if that's an acceptable case. + * Last line will not end with '\n'; this is to facilitate the usual case of + * string compares with a single line of output. + * @param pathname The file to read. + * @param desc Optional description of the file for error message; + * defaults to "in <pathname>" + */ +static std::string readfile(const std::string& pathname, const std::string& desc="") +{ + std::string use_desc(desc); + if (use_desc.empty()) + { + use_desc = STRINGIZE("in " << pathname); + } + std::ifstream inf(pathname.c_str()); + std::string output; + tut::ensure(STRINGIZE("No output " << use_desc), std::getline(inf, output)); + std::string more; + while (std::getline(inf, more)) + { + output += '\n' + more; + } + return output; +} + +/** + * Construct an LLProcess to run a Python script. + */ +struct PythonProcessLauncher +{ + /** + * @param desc Arbitrary description for error messages + * @param script Python script, any form acceptable to NamedTempFile, + * typically either a std::string or an expression of the form + * (lambda::_1 << "script content with " << variable_data) + */ + template <typename CONTENT> + PythonProcessLauncher(const std::string& desc, const CONTENT& script): + mDesc(desc), + mScript("py", script) + { + const char* PYTHON(getenv("PYTHON")); + tut::ensure("Set $PYTHON to the Python interpreter", PYTHON); + + mParams["executable"] = PYTHON; + mParams["args"].append(mScript.getName()); + } + + /// Run Python script and wait for it to complete. + void run() + { + mPy = LLProcess::create(mParams); + tut::ensure(STRINGIZE("Couldn't launch " << mDesc << " script"), mPy); + // One of the irritating things about LLProcess is that + // there's no API to wait for the child to terminate -- but given + // its use in our graphics-intensive interactive viewer, it's + // understandable. + while (mPy->isRunning()) + { + sleep(1); + } + } + + /** + * Run a Python script using LLProcess, expecting that it will + * write to the file passed as its sys.argv[1]. Retrieve that output. + * + * Until January 2012, LLProcess provided distressingly few + * mechanisms for a child process to communicate back to its caller -- + * not even its return code. We've introduced a convention by which we + * create an empty temp file, pass the name of that file to our child + * as sys.argv[1] and expect the script to write its output to that + * file. This function implements the C++ (parent process) side of + * that convention. + */ + std::string run_read() + { + NamedTempFile out("out", ""); // placeholder + // pass name of this temporary file to the script + mParams["args"].append(out.getName()); + run(); + // assuming the script wrote to that file, read it + return readfile(out.getName(), STRINGIZE("from " << mDesc << " script")); + } + + LLSD mParams; + LLProcessPtr mPy; + std::string mDesc; + NamedTempFile mScript; +}; + +/// convenience function for PythonProcessLauncher::run() +template <typename CONTENT> +static void python(const std::string& desc, const CONTENT& script) +{ + PythonProcessLauncher py(desc, script); + py.run(); +} + +/// convenience function for PythonProcessLauncher::run_read() +template <typename CONTENT> +static std::string python_out(const std::string& desc, const CONTENT& script) +{ + PythonProcessLauncher py(desc, script); + return py.run_read(); +} + +/// Create a temporary directory and clean it up later. +class NamedTempDir: public boost::noncopyable +{ +public: + // Use python() function to create a temp directory: I've found + // nothing in either Boost.Filesystem or APR quite like Python's + // tempfile.mkdtemp(). + // Special extra bonus: on Mac, mkdtemp() reports a pathname + // starting with /var/folders/something, whereas that's really a + // symlink to /private/var/folders/something. Have to use + // realpath() to compare properly. + NamedTempDir(): + mPath(python_out("mkdtemp()", + "from __future__ import with_statement\n" + "import os.path, sys, tempfile\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write(os.path.realpath(tempfile.mkdtemp()))\n")) + {} + + ~NamedTempDir() + { + aprchk(apr_dir_remove(mPath.c_str(), gAPRPoolp)); + } + + std::string getName() const { return mPath; } + +private: + std::string mPath; +}; + +/***************************************************************************** +* TUT +*****************************************************************************/ +namespace tut +{ + struct llprocess_data + { + LLAPRPool pool; + }; + typedef test_group<llprocess_data> llprocess_group; + typedef llprocess_group::object object; + llprocess_group llprocessgrp("llprocess"); + + struct Item + { + Item(): tries(0) {} + unsigned tries; + std::string which; + std::string what; + }; + +/*==========================================================================*| +#define tabent(symbol) { symbol, #symbol } + static struct ReasonCode + { + int code; + const char* name; + } reasons[] = + { + tabent(APR_OC_REASON_DEATH), + tabent(APR_OC_REASON_UNWRITABLE), + tabent(APR_OC_REASON_RESTART), + tabent(APR_OC_REASON_UNREGISTER), + tabent(APR_OC_REASON_LOST), + tabent(APR_OC_REASON_RUNNING) + }; +#undef tabent +|*==========================================================================*/ + + struct WaitInfo + { + WaitInfo(apr_proc_t* child_): + child(child_), + rv(-1), // we haven't yet called apr_proc_wait() + rc(0), + why(apr_exit_why_e(0)) + {} + apr_proc_t* child; // which subprocess + apr_status_t rv; // return from apr_proc_wait() + int rc; // child's exit code + apr_exit_why_e why; // APR_PROC_EXIT, APR_PROC_SIGNAL, APR_PROC_SIGNAL_CORE + }; + + void child_status_callback(int reason, void* data, int status) + { +/*==========================================================================*| + std::string reason_str; + BOOST_FOREACH(const ReasonCode& rcp, reasons) + { + if (reason == rcp.code) + { + reason_str = rcp.name; + break; + } + } + if (reason_str.empty()) + { + reason_str = STRINGIZE("unknown reason " << reason); + } + std::cout << "child_status_callback(" << reason_str << ")\n"; +|*==========================================================================*/ + + if (reason == APR_OC_REASON_DEATH || reason == APR_OC_REASON_LOST) + { + // Somewhat oddly, APR requires that you explicitly unregister + // even when it already knows the child has terminated. + apr_proc_other_child_unregister(data); + + WaitInfo* wi(static_cast<WaitInfo*>(data)); + // It's just wrong to call apr_proc_wait() here. The only way APR + // knows to call us with APR_OC_REASON_DEATH is that it's already + // reaped this child process, so calling wait() will only produce + // "huh?" from the OS. We must rely on the status param passed in, + // which unfortunately comes straight from the OS wait() call. +// wi->rv = apr_proc_wait(wi->child, &wi->rc, &wi->why, APR_NOWAIT); + wi->rv = APR_CHILD_DONE; // fake apr_proc_wait() results +#if defined(LL_WINDOWS) + wi->why = APR_PROC_EXIT; + wi->rc = status; // no encoding on Windows (no signals) +#else // Posix + if (WIFEXITED(status)) + { + wi->why = APR_PROC_EXIT; + wi->rc = WEXITSTATUS(status); + } + else if (WIFSIGNALED(status)) + { + wi->why = APR_PROC_SIGNAL; + wi->rc = WTERMSIG(status); + } + else // uh, shouldn't happen? + { + wi->why = APR_PROC_EXIT; + wi->rc = status; // someone else will have to decode + } +#endif // Posix + } + } + + template<> template<> + void object::test<1>() + { + set_test_name("raw APR nonblocking I/O"); + + // Create a script file in a temporary place. + NamedTempFile script("py", + "import sys" EOL + "import time" EOL + EOL + "time.sleep(2)" EOL + "print >>sys.stdout, 'stdout after wait'" EOL + "sys.stdout.flush()" EOL + "time.sleep(2)" EOL + "print >>sys.stderr, 'stderr after wait'" EOL + "sys.stderr.flush()" EOL + ); + + // Arrange to track the history of our interaction with child: what we + // fetched, which pipe it came from, how many tries it took before we + // got it. + std::vector<Item> history; + history.push_back(Item()); + + // Run the child process. + apr_procattr_t *procattr = NULL; + aprchk(apr_procattr_create(&procattr, pool.getAPRPool())); + aprchk(apr_procattr_io_set(procattr, APR_CHILD_BLOCK, APR_CHILD_BLOCK, APR_CHILD_BLOCK)); + aprchk(apr_procattr_cmdtype_set(procattr, APR_PROGRAM_PATH)); + + std::vector<const char*> argv; + apr_proc_t child; + argv.push_back("python"); + // Have to have a named copy of this std::string so its c_str() value + // will persist. + std::string scriptname(script.getName()); + argv.push_back(scriptname.c_str()); + argv.push_back(NULL); + + aprchk(apr_proc_create(&child, argv[0], + &argv[0], + NULL, // if we wanted to pass explicit environment + procattr, + pool.getAPRPool())); + + // We do not want this child process to outlive our APR pool. On + // destruction of the pool, forcibly kill the process. Tell APR to try + // SIGTERM and wait 3 seconds. If that didn't work, use SIGKILL. + apr_pool_note_subprocess(pool.getAPRPool(), &child, APR_KILL_AFTER_TIMEOUT); + + // arrange to call child_status_callback() + WaitInfo wi(&child); + apr_proc_other_child_register(&child, child_status_callback, &wi, child.in, pool.getAPRPool()); + + // TODO: + // Stuff child.in until it (would) block to verify EWOULDBLOCK/EAGAIN. + // Have child script clear it later, then write one more line to prove + // that it gets through. + + // Monitor two different output pipes. Because one will be closed + // before the other, keep them in a list so we can drop whichever of + // them is closed first. + typedef std::pair<std::string, apr_file_t*> DescFile; + typedef std::list<DescFile> DescFileList; + DescFileList outfiles; + outfiles.push_back(DescFile("out", child.out)); + outfiles.push_back(DescFile("err", child.err)); + + while (! outfiles.empty()) + { + // This peculiar for loop is designed to let us erase(dfli). With + // a list, that invalidates only dfli itself -- but even so, we + // lose the ability to increment it for the next item. So at the + // top of every loop, while dfli is still valid, increment + // dflnext. Then before the next iteration, set dfli to dflnext. + for (DescFileList::iterator + dfli(outfiles.begin()), dflnext(outfiles.begin()), dflend(outfiles.end()); + dfli != dflend; dfli = dflnext) + { + // Only valid to increment dflnext once we're sure it's not + // already at dflend. + ++dflnext; + + char buf[4096]; + + apr_status_t rv = apr_file_gets(buf, sizeof(buf), dfli->second); + if (APR_STATUS_IS_EOF(rv)) + { +// std::cout << "(EOF on " << dfli->first << ")\n"; +// history.back().which = dfli->first; +// history.back().what = "*eof*"; +// history.push_back(Item()); + outfiles.erase(dfli); + continue; + } + if (rv == EWOULDBLOCK || rv == EAGAIN) + { +// std::cout << "(waiting; apr_file_gets(" << dfli->first << ") => " << rv << ": " << manager.strerror(rv) << ")\n"; + ++history.back().tries; + continue; + } + aprchk_("apr_file_gets(buf, sizeof(buf), dfli->second)", rv); + // Is it even possible to get APR_SUCCESS but read 0 bytes? + // Hope not, but defend against that anyway. + if (buf[0]) + { +// std::cout << dfli->first << ": " << buf; + history.back().which = dfli->first; + history.back().what.append(buf); + if (buf[strlen(buf) - 1] == '\n') + history.push_back(Item()); + else + { + // Just for pretty output... if we only read a partial + // line, terminate it. +// std::cout << "...\n"; + } + } + } + // Do this once per tick, as we expect the viewer will + apr_proc_other_child_refresh_all(APR_OC_REASON_RUNNING); + sleep(1); + } + apr_file_close(child.in); + apr_file_close(child.out); + apr_file_close(child.err); + + // Okay, we've broken the loop because our pipes are all closed. If we + // haven't yet called wait, give the callback one more chance. This + // models the fact that unlike this small test program, the viewer + // will still be running. + if (wi.rv == -1) + { + std::cout << "last gasp apr_proc_other_child_refresh_all()\n"; + apr_proc_other_child_refresh_all(APR_OC_REASON_RUNNING); + } + + if (wi.rv == -1) + { + std::cout << "child_status_callback(APR_OC_REASON_DEATH) wasn't called" << std::endl; + wi.rv = apr_proc_wait(wi.child, &wi.rc, &wi.why, APR_NOWAIT); + } +// std::cout << "child done: rv = " << rv << " (" << manager.strerror(rv) << "), why = " << why << ", rc = " << rc << '\n'; + aprchk_("apr_proc_wait(wi->child, &wi->rc, &wi->why, APR_NOWAIT)", wi.rv, APR_CHILD_DONE); + ensure_equals_(wi.why, APR_PROC_EXIT); + ensure_equals_(wi.rc, 0); + + // Beyond merely executing all the above successfully, verify that we + // obtained expected output -- and that we duly got control while + // waiting, proving the non-blocking nature of these pipes. + try + { + unsigned i = 0; + ensure("blocking I/O on child pipe (0)", history[i].tries); + ensure_equals_(history[i].which, "out"); + ensure_equals_(history[i].what, "stdout after wait" EOL); +// ++i; +// ensure_equals_(history[i].which, "out"); +// ensure_equals_(history[i].what, "*eof*"); + ++i; + ensure("blocking I/O on child pipe (1)", history[i].tries); + ensure_equals_(history[i].which, "err"); + ensure_equals_(history[i].what, "stderr after wait" EOL); +// ++i; +// ensure_equals_(history[i].which, "err"); +// ensure_equals_(history[i].what, "*eof*"); + } + catch (const failure&) + { + std::cout << "History:\n"; + BOOST_FOREACH(const Item& item, history) + { + std::string what(item.what); + if ((! what.empty()) && what[what.length() - 1] == '\n') + { + what.erase(what.length() - 1); + if ((! what.empty()) && what[what.length() - 1] == '\r') + { + what.erase(what.length() - 1); + what.append("\\r"); + } + what.append("\\n"); + } + std::cout << " " << item.which << ": '" << what << "' (" + << item.tries << " tries)\n"; + } + std::cout << std::flush; + // re-raise same error; just want to enrich the output + throw; + } + } + + template<> template<> + void object::test<2>() + { + set_test_name("setWorkingDirectory()"); + // We want to test setWorkingDirectory(). But what directory is + // guaranteed to exist on every machine, under every OS? Have to + // create one. Naturally, ensure we clean it up when done. + NamedTempDir tempdir; + PythonProcessLauncher py("getcwd()", + "from __future__ import with_statement\n" + "import os, sys\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write(os.getcwd())\n"); + // Before running, call setWorkingDirectory() + py.mParams["cwd"] = tempdir.getName(); + ensure_equals("os.getcwd()", py.run_read(), tempdir.getName()); + } + + template<> template<> + void object::test<3>() + { + set_test_name("arguments"); + PythonProcessLauncher py("args", + "from __future__ import with_statement\n" + "import sys\n" + // note nonstandard output-file arg! + "with open(sys.argv[3], 'w') as f:\n" + " for arg in sys.argv[1:]:\n" + " print >>f, arg\n"); + // We expect that PythonProcessLauncher has already appended + // its own NamedTempFile to mParams["args"] (sys.argv[0]). + py.mParams["args"].append("first arg"); // sys.argv[1] + py.mParams["args"].append("second arg"); // sys.argv[2] + // run_read() appends() one more argument, hence [3] + std::string output(py.run_read()); + boost::split_iterator<std::string::const_iterator> + li(output, boost::first_finder("\n")), lend; + ensure("didn't get first arg", li != lend); + std::string arg(li->begin(), li->end()); + ensure_equals(arg, "first arg"); + ++li; + ensure("didn't get second arg", li != lend); + arg.assign(li->begin(), li->end()); + ensure_equals(arg, "second arg"); + ++li; + ensure("didn't get output filename?!", li != lend); + arg.assign(li->begin(), li->end()); + ensure("output filename empty?!", ! arg.empty()); + ++li; + ensure("too many args", li == lend); + } + + template<> template<> + void object::test<4>() + { + set_test_name("explicit kill()"); + PythonProcessLauncher py("kill()", + "from __future__ import with_statement\n" + "import sys, time\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('ok')\n" + "# now sleep; expect caller to kill\n" + "time.sleep(120)\n" + "# if caller hasn't managed to kill by now, bad\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('bad')\n"); + NamedTempFile out("out", "not started"); + py.mParams["args"].append(out.getName()); + py.mPy = LLProcess::create(py.mParams); + ensure("couldn't launch kill() script", py.mPy); + // Wait for the script to wake up and do its first write + int i = 0, timeout = 60; + for ( ; i < timeout; ++i) + { + sleep(1); + if (readfile(out.getName(), "from kill() script") == "ok") + break; + } + // If we broke this loop because of the counter, something's wrong + ensure("script never started", i < timeout); + // script has performed its first write and should now be sleeping. + py.mPy->kill(); + // wait for the script to terminate... one way or another. + while (py.mPy->isRunning()) + { + sleep(1); + } + // If kill() failed, the script would have woken up on its own and + // overwritten the file with 'bad'. But if kill() succeeded, it should + // not have had that chance. + ensure_equals("kill() script output", readfile(out.getName()), "ok"); + } + + template<> template<> + void object::test<5>() + { + set_test_name("implicit kill()"); + NamedTempFile out("out", "not started"); + LLProcess::id pid(0); + { + PythonProcessLauncher py("kill()", + "from __future__ import with_statement\n" + "import sys, time\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('ok')\n" + "# now sleep; expect caller to kill\n" + "time.sleep(120)\n" + "# if caller hasn't managed to kill by now, bad\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('bad')\n"); + py.mParams["args"].append(out.getName()); + py.mPy = LLProcess::create(py.mParams); + ensure("couldn't launch kill() script", py.mPy); + // Capture id for later + pid = py.mPy->getProcessID(); + // Wait for the script to wake up and do its first write + int i = 0, timeout = 60; + for ( ; i < timeout; ++i) + { + sleep(1); + if (readfile(out.getName(), "from kill() script") == "ok") + break; + } + // If we broke this loop because of the counter, something's wrong + ensure("script never started", i < timeout); + // Script has performed its first write and should now be sleeping. + // Destroy the LLProcess, which should kill the child. + } + // wait for the script to terminate... one way or another. + while (LLProcess::isRunning(pid)) + { + sleep(1); + } + // If kill() failed, the script would have woken up on its own and + // overwritten the file with 'bad'. But if kill() succeeded, it should + // not have had that chance. + ensure_equals("kill() script output", readfile(out.getName()), "ok"); + } + + template<> template<> + void object::test<6>() + { + set_test_name("autokill"); + NamedTempFile from("from", "not started"); + NamedTempFile to("to", ""); + LLProcess::id pid(0); + { + PythonProcessLauncher py("autokill", + "from __future__ import with_statement\n" + "import sys, time\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('ok')\n" + "# wait for 'go' from test program\n" + "for i in xrange(60):\n" + " time.sleep(1)\n" + " with open(sys.argv[2]) as f:\n" + " go = f.read()\n" + " if go == 'go':\n" + " break\n" + "else:\n" + " with open(sys.argv[1], 'w') as f:\n" + " f.write('never saw go')\n" + " sys.exit(1)\n" + "# okay, saw 'go', write 'ack'\n" + "with open(sys.argv[1], 'w') as f:\n" + " f.write('ack')\n"); + py.mParams["args"].append(from.getName()); + py.mParams["args"].append(to.getName()); + py.mParams["autokill"] = false; + py.mPy = LLProcess::create(py.mParams); + ensure("couldn't launch kill() script", py.mPy); + // Capture id for later + pid = py.mPy->getProcessID(); + // Wait for the script to wake up and do its first write + int i = 0, timeout = 60; + for ( ; i < timeout; ++i) + { + sleep(1); + if (readfile(from.getName(), "from autokill script") == "ok") + break; + } + // If we broke this loop because of the counter, something's wrong + ensure("script never started", i < timeout); + // Now destroy the LLProcess, which should NOT kill the child! + } + // If the destructor killed the child anyway, give it time to die + sleep(2); + // How do we know it's not terminated? By making it respond to + // a specific stimulus in a specific way. + { + std::ofstream outf(to.getName().c_str()); + outf << "go"; + } // flush and close. + // now wait for the script to terminate... one way or another. + while (LLProcess::isRunning(pid)) + { + sleep(1); + } + // If the LLProcess destructor implicitly called kill(), the + // script could not have written 'ack' as we expect. + ensure_equals("autokill script output", readfile(from.getName()), "ack"); + } +} // namespace tut |
