chromium//src/content/browser/loader/cross_site_document_resource_handler_unittest.cc

    // Content-Type header.
    std::string charset;
    bool had_charset;
    std::string boundary;
    response_headers->AddHeader(std::string("Content-Type: ") +
                                response_mime_type);
    response->head.mime_type = response_mime_type;
    net::HttpUtil::ParseContentType(response_mime_type,
                                    &response->head.mime_type, &charset,
                                    &had_charset, &boundary);


Added in https://chromium.googlesource.com/chromium/src/+/0753a97df25d497e872e2945848907ca9a6661bf



chromium//src/net/http/http_util.cc

  if ((!eq && *had_charset) || type_has_charset) {
    *had_charset = true;
    *charset = base::ToLowerASCII(charset_value);
  }

Touched in https://chromium.googlesource.com/chromium/src/+/c0399701530ae196c656e644d19646e88cfa58af but has been around since initial.commit.

That should probably read

  if ((!eq && had_charset) || type_has_charset) {
    *had_charset = true;
    *charset = base::ToLowerASCII(charset_value);
  }


So it's a true positive and I guess msan progressed to find it (?)



bnc, you probably know that code -- want to fix?

...or maybe the test just needs to initialize had_charset. I don't understand the intent of that check in ParseContentType.

https://chromiumcodereview.appspot.com/10834190 suggests that the bool just needs to be initialized and that this is a confusing API.

Every caller passes in false as initial value as far as I can tell, so maybe the first part of the conditional can just be removed?

The only thing that looks at the output value is the test.

However, for this loop removing the read would be a change in behavior:

https://cs.chromium.org/chromium/src/net/http/http_response_headers.cc?l=857

  bool had_charset = false;

  size_t iter = 0;
  while (EnumerateHeader(&iter, name, &value))
    HttpUtil::ParseContentType(value, mime_type, charset, &had_charset, NULL);


So the code probably has to stay as-is (but its header comment should point out that *had_charset is always read and must be initialized, and the cc comments should be clearer on what's happening there since I didn't understand it) -- bnc, want to do this part?

And the value needs to be initialized in the test. lukasza, I'll send you a CL for that.

Yes I'll do the //net part.  As for CrossSiteDocumentResourceHandlerTest, I propose https://crrev.com/c/1233814 instead.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8f2ce189b1622824fc2ec7c83df01443b495f65a

commit 8f2ce189b1622824fc2ec7c83df01443b495f65a
Author: Nico Weber <thakis@chromium.org>
Date: Wed Sep 19 14:41:37 2018

Fix an uninitialized read in a test found by MSan.

A different fix is under discussion at
https://chromium-review.googlesource.com/c/chromium/src/+/1233814
but since there's some disagreement on the approach, let's land
this while the review over there is ongoing, to heal the bot.

Bug:  886629 
Change-Id: I80c64f6bf969b5b1edf108d2a46155f2f4252595
Reviewed-on: https://chromium-review.googlesource.com/1232591
Reviewed-by: Matt Menke <mmenke@chromium.org>
Reviewed-by: Łukasz Anforowicz <lukasza@chromium.org>
Commit-Queue: Nico Weber <thakis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#592393}
[modify] https://crrev.com/8f2ce189b1622824fc2ec7c83df01443b495f65a/content/browser/loader/cross_site_document_resource_handler_unittest.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/579a4f9ae5c3337c2f8d2605f669d6719efb4e97

commit 579a4f9ae5c3337c2f8d2605f669d6719efb4e97
Author: Bence Béky <bnc@chromium.org>
Date: Wed Sep 19 15:34:52 2018

Fix Content-Type issues in CrossSiteDocumentResourceHandlerTest.

1. Remove |response->head.mime_type| assignment before
   ParseContentType call.
2. Rename Content-Type test parameter to response_content_type.
3. Initialize |had_charset| to false to address MSAN bug;
   this will be removed from this CL as soon as
   https://crrev.com/c/1232591 lands and I rebase.

Bug:  886629 
Change-Id: I32fbe9d5c1fa6a0ba735bfd335923f0e1a723b04
Reviewed-on: https://chromium-review.googlesource.com/1233814
Reviewed-by: Łukasz Anforowicz <lukasza@chromium.org>
Reviewed-by: Matt Menke <mmenke@chromium.org>
Commit-Queue: Bence Béky <bnc@chromium.org>
Cr-Commit-Position: refs/heads/master@{#592410}
[modify] https://crrev.com/579a4f9ae5c3337c2f8d2605f669d6719efb4e97/content/browser/loader/cross_site_document_resource_handler_unittest.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/45844e36bdaa5664ea63467bbe300989798e14af

commit 45844e36bdaa5664ea63467bbe300989798e14af
Author: Bence Béky <bnc@chromium.org>
Date: Wed Sep 19 18:53:37 2018

Fix HttpUtil::ParseContentType() API comment.

Bug:  886629 
Change-Id: Ie0f096552bec89f57041caebcf248da76f3a468d
Reviewed-on: https://chromium-review.googlesource.com/1233936
Commit-Queue: Bence Béky <bnc@chromium.org>
Reviewed-by: Matt Menke <mmenke@chromium.org>
Cr-Commit-Position: refs/heads/master@{#592496}
[modify] https://crrev.com/45844e36bdaa5664ea63467bbe300989798e14af/net/http/http_util.h

https://ci.chromium.org/buildbot/chromium.clang/ToTLinuxMSan/ will hopefully cycle green. Thanks!