Do you have a stack or something?

+ folks from https://codereview.chromium.org/2737783003

This is a symbolized stack trace I get when running
out/linux_msan/browser_test --gtest_filter=PrintPreviewDest*

Uninitialized bytes in __interceptor_send at offset 5 inside [0x71d0000a0030, 151)
==13213==WARNING: MemorySanitizer: use-of-uninitialized-value
    #0 0x7ff83c60ce06 in gnutls_record_check_pending ??:?
    #1 0x7ff83c60ce06 in ?? ??:0
    #2 0x7ff83c60d14e in gnutls_record_check_pending ??:?
    #3 0x7ff83c60d14e in ?? ??:0
    #4 0x7ff83c60df0e in gnutls_transport_set_errno_function ??:?
    #5 0x7ff83c60df0e in ?? ??:0
    #6 0x7ff83c611eca in gnutls_handshake_set_max_packet_length ??:?
    #7 0x7ff83c611eca in ?? ??:0
    #8 0x7ff83c612204 in gnutls_handshake_set_max_packet_length ??:?
    #9 0x7ff83c612204 in ?? ??:0
    #10 0x7ff83c6135b4 in gnutls_handshake ??:0:0
    #11 0x7ff844994fb5 in http_setup_ssl /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:5177:20
    #12 0x7ff844997c09 in _httpUpdate /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:3077:11
    #13 0x7ff844988d7f in httpUpdate /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:3239:10
    #14 0x7ff844988d7f in http_upgrade /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:5598:0
    #15 0x7ff8449868fb in httpReconnect2 /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:2653:13
    #16 0x7ff844986077 in httpConnect2 /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/http.c:496:21
    #17 0x7ff844a31d3b in _cupsConnect /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/request.c:1047:21
    #18 0x7ff844a30bb7 in cupsDoIORequest /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/request.c:159:17
    #19 0x7ff8449433af in _cupsGetDests /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/dest.c:1501:19
    #20 0x7ff84494ad38 in cupsGetDests2 /home/tom/chromium/src/third_party/instrumented_libraries/scripts/out/Instrumented-msan-chained-origins-trusty/gen/third_party/instrumented_libraries/libcups2/cups-1.7.2/cups/dest.c:1725:15
    #21 0x1491e51a in printing::PrintBackendCUPS::GetDests(cups_dest_s**) printing/backend/print_backend_cups.cc:227:12
    #22 0x1491f7c0 in printing::PrintBackendCUPS::GetDefaultPrinterName() printing/backend/print_backend_cups.cc:112:19
    #23 0x16e0c5f4 in printing::(anonymous namespace)::GetDefaultPrinterOnBlockingPoolThread() chrome/browser/ui/webui/print_preview/printer_backend_proxy.cc:62:48
    #24 0x16e0c9ed in Invoke<> base/bind_internal.h:164:12
    #25 0x16e0c9ed in MakeItSo<std::__1::basic_string<char> (*const &)()> base/bind_internal.h:285:0
    #26 0x16e0c9ed in RunImpl<std::__1::basic_string<char> (*const &)(), const std::__1::tuple<> &> base/bind_internal.h:361:0
    #27 0x16e0c9ed in base::internal::Invoker<base::internal::BindState<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > (*)()>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > ()>::Run(base::internal::BindStateBase*) base/bind_internal.h:339:0
    #28 0x4c90669 in Run base/callback.h:91:12
    #29 0x4c90669 in void base::internal::ReturnAsParamAdapter<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >(base::Callback<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > (), (base::internal::CopyMode)0, (base::internal::RepeatMode)0>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >*) base/post_task_and_reply_with_result_internal.h:20:0

Maybe it's instrumented and we just don't have symbols for us? The relevant bits are probably this:

    #21 0x1491e51a in printing::PrintBackendCUPS::GetDests(cups_dest_s**) printing/backend/print_backend_cups.cc:227:12
    #22 0x1491f7c0 in printing::PrintBackendCUPS::GetDefaultPrinterName() printing/backend/print_backend_cups.cc:112:19
    #23 0x16e0c5f4 in printing::(anonymous namespace)::GetDefaultPrinterOnBlockingPoolThread() chrome/browser/ui/webui/print_preview/printer_backend_proxy.cc:62:48


Do you have more output? Usually you get several stacks: "uninitialized memory used here; first created here"

It's not in the list under third_party/instrumented_libraries.
An unsymbolized stack trace contains library paths where it is easy to see whether the library is loaded from the build directory, or from somewhere like /usr/lib.

Doesn't msan depend on everything being instrumented? It sounds like cups has depended on gnutls for a long time, why have we never noticed that this one is missing?

Maybe this code never actually executed? On the other hand, we configure cups with --enable-gnutls.

Sorry, left out the last bit, it's blaming gnutls but I'm not sure if this is an actual bug or false positive.

  Uninitialized value was stored to memory at
    #0 0x733c1b in __msan_memcpy ??:0:0
    #1 0x7ff83c60b200 in gnutls_compression_list ??:?
    #2 0x7ff83c60b200 in ?? ??:0

  Uninitialized value was created by a heap allocation
    #0 0x73ad6d in __interceptor_malloc ??:0:0
    #1 0x7ff83c60bee8 in gnutls_compression_list ??:?
    #2 0x7ff83c60bee8 in ?? ??:0

I took a look at the code it's blaming, and it's copying an array of constants.

If you run 'LD_LIBRARY_PATH=/path/to/instrumented/libs ldd out/Release/chrome', there shouldn't be any dependencies on system libraries.  The gnutls dependency is definitely a bug.

thakis@ are you actively working on this?  If you assign this to me, I can get it done within the next few days

I won't say no to that :-)

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e80386955278208db321641577837ed7d1771172

commit e80386955278208db321641577837ed7d1771172
Author: thomasanderson <thomasanderson@google.com>
Date: Fri Apr 28 07:45:38 2017

Add libgnutls26 to instrumented libraries

BUG= 716118 
R=eugenis@chromium.org

Review-Url: https://codereview.chromium.org/2851493005
Cr-Commit-Position: refs/heads/master@{#467923}

[modify] https://crrev.com/e80386955278208db321641577837ed7d1771172/third_party/instrumented_libraries/BUILD.gn
[modify] https://crrev.com/e80386955278208db321641577837ed7d1771172/third_party/instrumented_libraries/scripts/install-build-deps.sh

gnutls is pretty buggy, it's a bit frightening that we depend on it. 

https://bugs.chromium.org/p/oss-fuzz/issues/list?can=1&q=status%3AFixed%2CVerified+Type%3ABug%2CBug-Security+-component%3AInfra+gnutls&colspec=ID+Type+Component+Status+Proj+Reported+Owner+Summary&cells=ids

Shouldn't we try to get rid of this dependency? 

PrintPreviewDestinationSearchTest.Select seems to be failing again on the linux msan bot. Here's the first failure: https://build.chromium.org/p/chromium.memory/builders/Linux%20MSan%20Tests/builds/411 .

I'm looking into why it's failing.  I reverted the previous cause of the failure. I'm going to track the test failure separately  crbug.com/716595 

Does libcups need to be recompiled with different flags?  I'm still seeing calls into /usr/lib/x86_64-linux-gnu/libgnutls.so.26+0x1ae06

https://codereview.chromium.org/2847033002/ hasn't landed yet.  You can apply this patch locally to test.  Make sure to run "GYP_DEFINES='msan=1 use_prebuilt_instrumented_libraries=1' gclient runhooks" first.

It looks like PrintPreviewDestinationSearchTest is still failing with libgnutls, but there's no msan errors now.

Also, now a new test is failing with libgnutls added:
https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Ftryserver.chromium.linux%2Flinux_chromium_msan_rel_ng%2F627%2F%2B%2Frecipes%2Fsteps%2Fnet_unittests__with_patch_%2F0%2Flogs%2FNSS__x2f_ClientCertStoreTest__x2f_0.CertAuthorityFiltering%2F0

Yes.  It looks like there's a different problem with PrintPreviewDestinationSearchTest.  Not sure why it's been passing, but I've got a fix.

skau@ feel like debugging NSS/ClientCertStoreTest/0.CertAuthorityFiltering too? :)

Why not. I'll take a look.  I have it patched already.

AFAICT, there might be a problem with the nss libarary we're linking to.  IsIssuedByEncoded is implemented with CERT_CompareName https://cs.chromium.org/chromium/src/net/cert/x509_util_nss.cc?l=393 which comes from the nss.h header.

Just waiting for bug 716594 to be fixed so that I can land https://codereview.chromium.org/2847033002/

Is bug 716594 actually blocking this? NSS/ClientCertStoreTest/CertAuthorityFiltering 
 was disabled in https://codereview.chromium.org/2850083002.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8a38359e077609f8063d4354f43a32864b561ddd

commit 8a38359e077609f8063d4354f43a32864b561ddd
Author: thomasanderson <thomasanderson@google.com>
Date: Wed May 03 06:22:27 2017

Roll msan instrumented libraries

This CL rolls the instrumented libraries after the addition of
libgnutls26 in https://codereview.chromium.org/2851493005/

CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_chromium_msan_rel_ng;master.tryserver.chromium.linux:linux_chromium_chromeos_msan_rel_ng
BUG= 716118 
R=eugenis@chromium.org

Review-Url: https://codereview.chromium.org/2847033002
Cr-Commit-Position: refs/heads/master@{#468904}

[modify] https://crrev.com/8a38359e077609f8063d4354f43a32864b561ddd/third_party/instrumented_libraries/binaries/msan-chained-origins-trusty.tgz.sha1
[modify] https://crrev.com/8a38359e077609f8063d4354f43a32864b561ddd/third_party/instrumented_libraries/binaries/msan-no-origins-trusty.tgz.sha1