Upgrade chrome's OpenSSL to same version Android ships with.

Upgrade chrome's OpenSSL to same version Android ships with.

This change is basically from openssl.org with the patches applied as per patches/README in https://android.googlesource.com/platform/external/openssl, plus 3 patches for Chromium to pass compliation and tests. Please refer to b/5428019 for more details.

BUG=None
TEST=net_unittests
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=118977

[Johnny(Jianning) Ding, 2012-01-20 02:38:20]

On 2012/01/19 20:16:14, Johnny(Jianning) Ding wrote:
Addition information of this patch.

Like the patch description said, the is a openssl 1.0.0f with 
(1) patches need by Android(AOSP). See
https://android.googlesource.com/platform/external/openssl/patches. Notice that
currently AOSP's openssl is 1.0.0e, but Android main trunk is using 1.0.0f. Now
we use Android patches for 1.0.0f per according to communication with joth and
bdc.
(2) patches need by building openssl 1.0.0f on Linux desktop. They are
x509_hash_name_algorithm_change.patch,
STORE_ATTR_INFO_compare_declaration_conflict.patch(better name it to "using
local store.h"?) and empty_OPENSSL_cpuid_setup.patch. All details are in
README.chromium.

Since Android's opensslconf.h was changed, upgraded the
config/android/openssl/opensslconf.h accordingly.

openssl.gyp also changed according to changes from openssl 0.9.8o -> 1.0.0f

Above are all our changes I done based on the pure openssl 1.0.0f from
openssl.org

[joth, 2012-01-20 11:01:08]

couple quick comments. thanks very much for handling this

http://codereview.chromium.org/9254031/diff/4283/README.chromium
File README.chromium (right):

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode72
README.chromium:72: in current src directory instead of system-wide openssl
include path.
ouch. this sounds like the tip of a much deeper possible problem. I think a more
robust fix is to edit the .gyp file as necessary to ensure the local headers are
always going to be used in preference to any system wide openssl headers.

I'm OK doing it this way if we also raise a bug for this. (and link it here)

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode87
README.chromium:87: However the new openssl changed the has algorithm, Unless
you compile/install
nit: has->hash

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode98
README.chromium:98: 
this section is duplicated.

http://codereview.chromium.org/9254031/diff/4283/openssl.gyp
File openssl.gyp (right):

http://codereview.chromium.org/9254031/diff/4283/openssl.gyp#newcode122
openssl.gyp:122: 'openssl/crypto/asn1/p8_pkey.c',
nit: trailing w/s

http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
File patches/x509_hash_name_algorithm_change.patch (right):

http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
patches/x509_hash_name_algorithm_change.patch:20: +		h=hash_array[hash_index];
ouch. again a unfortunate side effect but this time because we're relying on the
system-wide SSL install to provide root certs. (in a certain format).

one alternative may be to re-configure our code to not use /etc/ssl/certs.. but
instead use a copy of the root certs we bundle up here in svn.
again, lets raise a new.crbug.com for that unless it's very easy to do.

[wtc, 2012-01-21 00:39:13]

jnd: I guess this CL is for the src/third_party/openssl directory
in the Chromium source tree?

I am not familiar with that copy of OpenSSL and how it is
used by Chromium, so I am not the best person to review this
CL.  Do you have any specific questions for me?

[Johnny(Jianning) Ding, 2012-01-21 16:54:31]

On 2012/01/21 00:39:13, wtc wrote:
> jnd: I guess this CL is for the src/third_party/openssl directory
> in the Chromium source tree?
> 
> I am not familiar with that copy of OpenSSL and how it is
> used by Chromium, so I am not the best person to review this
> CL.  Do you have any specific questions for me?
No more questions. Thanks anyway.

[Johnny(Jianning) Ding, 2012-01-23 14:20:56]

new patchset is uploading

http://codereview.chromium.org/9254031/diff/4283/README.chromium
File README.chromium (right):

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode72
README.chromium:72: in current src directory instead of system-wide openssl
include path.
On 2012/01/20 11:01:08, joth wrote:
> ouch. this sounds like the tip of a much deeper possible problem. I think a
more
> robust fix is to edit the .gyp file as necessary to ensure the local headers
are
> always going to be used in preference to any system wide openssl headers.
> 
> I'm OK doing it this way if we also raise a bug for this. (and link it here)

Done.

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode87
README.chromium:87: However the new openssl changed the has algorithm, Unless
you compile/install
On 2012/01/20 11:01:08, joth wrote:
> nit: has->hash 

Done.

http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode98
README.chromium:98: 
On 2012/01/20 11:01:08, joth wrote:
> this section is duplicated.

Done.

http://codereview.chromium.org/9254031/diff/4283/openssl.gyp
File openssl.gyp (right):

http://codereview.chromium.org/9254031/diff/4283/openssl.gyp#newcode122
openssl.gyp:122: 'openssl/crypto/asn1/p8_pkey.c',
On 2012/01/20 11:01:08, joth wrote:
> nit: trailing w/s

Done.

http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
File patches/x509_hash_name_algorithm_change.patch (right):

http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
patches/x509_hash_name_algorithm_change.patch:20: +		h=hash_array[hash_index];
On 2012/01/20 11:01:08, joth wrote:
> ouch. again a unfortunate side effect but this time because we're relying on
the
> system-wide SSL install to provide root certs. (in a certain format).
> 
> one alternative may be to re-configure our code to not use /etc/ssl/certs..
but
> instead use a copy of the root certs we bundle up here in svn.
> again, lets raise a http://new.crbug.com for that unless it's very easy to do.

Done.

[Johnny(Jianning) Ding, 2012-01-23 15:26:15]

patchset 4 and 5 were not completed due to too big change and unstable network
connection to app engine. Please wait for next patchset

On 2012/01/23 14:20:56, Johnny(Jianning) Ding wrote:
> new patchset is uploading
> 
> http://codereview.chromium.org/9254031/diff/4283/README.chromium
> File README.chromium (right):
> 
> http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode72
> README.chromium:72: in current src directory instead of system-wide openssl
> include path.
> On 2012/01/20 11:01:08, joth wrote:
> > ouch. this sounds like the tip of a much deeper possible problem. I think a
> more
> > robust fix is to edit the .gyp file as necessary to ensure the local headers
> are
> > always going to be used in preference to any system wide openssl headers.
> > 
> > I'm OK doing it this way if we also raise a bug for this. (and link it here)
> 
> Done.
> 
> http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode87
> README.chromium:87: However the new openssl changed the has algorithm, Unless
> you compile/install
> On 2012/01/20 11:01:08, joth wrote:
> > nit: has->hash 
> 
> Done.
> 
> http://codereview.chromium.org/9254031/diff/4283/README.chromium#newcode98
> README.chromium:98: 
> On 2012/01/20 11:01:08, joth wrote:
> > this section is duplicated.
> 
> Done.
> 
> http://codereview.chromium.org/9254031/diff/4283/openssl.gyp
> File openssl.gyp (right):
> 
> http://codereview.chromium.org/9254031/diff/4283/openssl.gyp#newcode122
> openssl.gyp:122: 'openssl/crypto/asn1/p8_pkey.c',
> On 2012/01/20 11:01:08, joth wrote:
> > nit: trailing w/s
> 
> Done.
> 
>
http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
> File patches/x509_hash_name_algorithm_change.patch (right):
> 
>
http://codereview.chromium.org/9254031/diff/4283/patches/x509_hash_name_algor...
> patches/x509_hash_name_algorithm_change.patch:20: +		h=hash_array[hash_index];
> On 2012/01/20 11:01:08, joth wrote:
> > ouch. again a unfortunate side effect but this time because we're relying on
> the
> > system-wide SSL install to provide root certs. (in a certain format).
> > 
> > one alternative may be to re-configure our code to not use /etc/ssl/certs..
> but
> > instead use a copy of the root certs we bundle up here in svn.
> > again, lets raise a http://new.crbug.com for that unless it's very easy to
do.
> 
> Done.

[joth, 2012-01-24 11:02:39]

Looks great, except the patches/ directory is missing in patch set 7?

[joth, 2012-01-24 15:10:05]

LGTM

Thanks!