[openssl-users] 1.1.1 pre1 tests failing on Solaris SPARC
openssl-users at dukhovni.org
Wed Feb 21 16:55:06 UTC 2018
> On Feb 21, 2018, at 11:42 AM, Dennis Clarke <dclarke at blastwave.org> wrote:
> On 21/02/18 09:14 AM, Viktor Dukhovni wrote:
>>> On Feb 21, 2018, at 5:06 AM, Andy Polyakov <appro at openssl.org> wrote:
>>> I wonder how come the problem with asn1_encode_test.c went unnoticed so
>>> far. Objects on stack are customarily aligned at pointer size, even if
>>> their declaration doesn't imply corresponding guarantee. So there are
>>> two options here: a) it's first time it's tested with SPARC Solaris cc
>>> (note that it is regularly tested on SPARC Linux, naturally with gcc);
>>> b) compiler was recently patched/upgraded. Do note that I don't dispute
>>> suggested fix (or compiler's "right" to misalign buf in this case), only
>>> wonder how come it worked so far. Implied question would be what are
>>> other possible implications of b).
>> The code introduced the misaligned "bug" is master-only, added in Apr/2017,
>> so quite possibly nobody has ever built in SunOS+SPARC, in which case it
>> never worked, but simply was never tested until now.
> Pretty sure I have done builds and tests. In fact I am certain of it.
Were you testing "master"? Or OpenSSL_1_1_0-stable? The alignment of
"buf" is of course compiler-version dependent, and can also change from
unrelated later changes in the surrounding code. In any case, the problem
code is comparatively recent (less than a year old), and only users testing
the master development branch on SPARC would have run into the issue.
More information about the openssl-users