[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH v4 2/5] Network: Add regression tests for the command-line arguments



[snip]
>> --- /dev/null
>> +++ b/tests/networkxml2argvtest.c
>> @@ -0,0 +1,99 @@
>> +#include <config.h>
>> +
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <unistd.h>
>> +#include <string.h>
>> +
>> +#include <sys/types.h>
>> +#include <fcntl.h>
>> +
>> +#include "internal.h"
>> +#include "testutils.h"
>> +#include "network_conf.h"
>> +#include "command.h"
>> +#include "memory.h"
>> +#include "network/bridge_driver.h"
>> +
>> +static int testCompareXMLToArgvFiles(const char *inxml, const char *outargv) {
>> +    char *inXmlData = NULL;
>> +    char *outArgvData = NULL;
>> +    char *actual = NULL;
>> +    int ret = -1;
>> +    virNetworkDefPtr dev = NULL;
>> +    virNetworkObjPtr obj = NULL;
>> +    virCommandPtr cmd = NULL;
>> +    char *pidfile = NULL;
>> +
>> +    if (virtTestLoadFile(inxml, &inXmlData) < 0)
>> +        goto fail;
>> +
>> +    if (virtTestLoadFile(outargv, &outArgvData) < 0)
>> +        goto fail;
>> +
>> +    if (!(dev = virNetworkDefParseString(inXmlData)))
>> +        goto fail;
>> +
>> +    if (VIR_ALLOC(obj) < 0)
>> +        goto fail;
>> +
>> +    obj->def = dev;
>> +
>> +    if (networkBuildDhcpDaemonCommandLine(obj, &cmd, pidfile) != 0)
>> +        goto fail;
>> +
>> +    if (!(actual = virCommandToString(cmd)))
>> +        goto fail;
>> +
>> +    if (STRNEQ(outArgvData, actual)) {
>> +        virtTestDifference(stderr, outArgvData, actual);
>> +        goto fail;
>> +    }
>> +
>> +    ret = 0;
>> +
>> + fail:
>> +    free(actual);
>> +    VIR_FREE(pidfile);
>> +    virCommandFree(cmd);
>> +    virNetworkObjFree(obj);
>> +    return ret;
>> +}
>> +
>> +static int testCompareXMLToArgvHelper(const void *data) {
>> +    char inxml[PATH_MAX];
>> +    char outargv[PATH_MAX];
>> +    snprintf(inxml, PATH_MAX, "%s/networkxml2argvdata/%s.xml",
>> +             abs_srcdir, (const char*)data);
>> +    snprintf(outargv, PATH_MAX, "%s/networkxml2argvdata/%s.argv",
>> +             abs_srcdir, (const char*)data);
>> +    return testCompareXMLToArgvFiles(inxml, outargv);
>> +}
>> +
>> +
>> +static int
>> +mymain(void)
>> +{
>> +    int ret = 0;
>> +    char cwd[PATH_MAX];
>> +
>> +    abs_srcdir = getenv("abs_srcdir");
>> +    if (!abs_srcdir)
>> +        abs_srcdir = getcwd(cwd, sizeof(cwd));
>> +
>> +#define DO_TEST(name) \
>> +    if (virtTestRun("Network XML-2-Argv " name, \
>> +                    1, testCompareXMLToArgvHelper, (name)) < 0) \
>> +        ret = -1
>> +
>> +    DO_TEST("isolated-network");
>> +    DO_TEST("routed-network");
>> +    DO_TEST("nat-network");
>> +    DO_TEST("netboot-network");
>> +    DO_TEST("netboot-proxy-network");
>> +    DO_TEST("nat-network-dns-txt-record");
>> +
>> +    return (ret==0 ? EXIT_SUCCESS : EXIT_FAILURE);
>> +}
>> +
>> +VIRT_TEST_MAIN(mymain)
>   I like the idea of adding more tests but in my case this is crashing
> with a memory error. Running under valgrind I get:
>
>
> paphio:~/libvirt/tests -> valgrind ./networkxml2argvtest
> TEST: networkxml2argvtest
>       ..==1582== Source and destination overlap in memcpy(0x7feffc660,
> 0x7feffc660, 35)
> ==1582==    at 0x4A073BA: memcpy (mc_replace_strmem.c:602)
> ==1582==    by 0x4FC4F0: safe_copy (strerror_r.c:103)
> ==1582==    by 0x4FC5FC: rpl_strerror_r (strerror_r.c:155)
> ==1582==    by 0x4256A6: virStrerror (virterror.c:1239)
> ==1582==    by 0x42577A: virReportSystemErrorFull (virterror.c:1268)
> ==1582==    by 0x417C1E: hostsfileSave (dnsmasq.c:215)
> ==1582==    by 0x417EB8: dnsmasqSave (dnsmasq.c:313)
> ==1582==    by 0x409A2E: networkSaveDnsmasqHostsfile
> (bridge_driver.c:453)
> ==1582==    by 0x409FE1: networkBuildDnsmasqArgv (bridge_driver.c:603)
> ==1582==    by 0x40A23C: networkBuildDhcpDaemonCommandLine
> (bridge_driver.c:667)
> ==1582==    by 0x40765E: testCompareXMLToArgvFiles
> (networkxml2argvtest.c:42)
> ==1582==    by 0x4077AC: testCompareXMLToArgvHelper
> (networkxml2argvtest.c:70)
> ==1582==    by 0x407D0F: virtTestRun (testutils.c:134)
> ==1582==    by 0x407885: mymain (networkxml2argvtest.c:91)
> ==1582==    by 0x408ACA: virtTestMain (testutils.c:619)
> ==1582==    by 0x40795F: main (networkxml2argvtest.c:99)
> ==1582== 
> !..!==1582== Invalid free() / delete / delete[]
> ==1582==    at 0x4A05187: free (vg_replace_malloc.c:325)
> ==1582==    by 0x41C45E: virFree (memory.c:310)
> ==1582==    by 0x408ADD: virtTestMain (testutils.c:623)
> ==1582==    by 0x40795F: main (networkxml2argvtest.c:99)
> ==1582==  Address 0x7feffee50 is not stack'd, malloc'd or (recently)
> free'd
> ==1582== 
>                                    6   FAIL
> paphio:~/libvirt/tests -> 


I didn't run this under the valgrind however no tests were failing for
me. I guess I need to run under the valgrind and it may be connected to
the fact all of the tests were rewritten since the time of the previous
version of my patch so this may be connected however this was working
fine on my F-14 i386 box so I don't really know what's wrong on your
setup. I need to investigate it further.


>  So two of the tests are failing here, and each of them seems to lead to
> a different issue for valgrind. The basic problem is that we seems
> to end up in hostsfileSave() within dnsmasq.c with a system path:
>
> Breakpoint 1, hostsfileSave (hostsfile=0x865260) at util/dnsmasq.c:210
> 210 {
> (gdb) p *hostsfile
> $2 = {nhosts = 2, hosts = 0x862690, 
>   path = 0x862b40 "/var/lib/libvirt/dnsmasq/default.hostsfile"}
> (gdb) n
> 211     int err = hostsfileWrite(hostsfile->path, hostsfile->hosts,
> (gdb) n
> 214     if (err < 0) {
> (gdb) p err
> $3 = -13
> (gdb) 
>
>  and then in the end abs_srcdir_cleanup is set but when doing
>   VIR_FREE(abs_srcdir); in testutils.c:622 it explodes due to an invalid
> address...
>
>   I'm a bit puzzled about what's ahppening in the first error,
> so we call virReportSystemErrorFull, normal, then first thing it does is
>
>     char strerror_buf[1024];
>
>     const char *errnoDetail = virStrerror(theerrno, strerror_buf,
>                                           sizeof(strerror_buf));
>
> which then jump into strerror_r and other code from gnulib I guess
> and that fails with valgrind.
>  There is something fishy here, but what exactly ?!?
>
> Daniel
>
Well, don't know exactly however I did to rewrite this for the latest
codebase since test infrastructure changed since last version of my
patch. However the strange thing is that it's working fine on my F-14
i386 box. Need to investigate this further.

Michal

-- 
Michal Novotny <minovotn redhat com>, RHCE
Virtualization Team (xen userspace), Red Hat


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]