<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 02/26/2017 02:10 AM, Richard Levitte wrote:<br>
<blockquote
cite="mid:20170226.091035.1875080190946668032.levitte@openssl.org"
type="cite">
<pre wrap="">In message <a class="moz-txt-link-rfc2396E" href="mailto:a2e37ebc-183e-023d-f059-7b89d6e3a352@akamai.com"><a2e37ebc-183e-023d-f059-7b89d6e3a352@akamai.com></a> on Fri, 27 Jan 2017 10:54:35 -0600, Benjamin Kaduk via openssl-dev <a class="moz-txt-link-rfc2396E" href="mailto:openssl-dev@openssl.org"><openssl-dev@openssl.org></a> said:
openssl-dev> There was some discussion about 1.0.1 being EoL on a FreeBSD list [0],
openssl-dev> and whether it would make sense to move to 1.0.2 on their stable
openssl-dev> branch, which led to someone making the claim that 1.0.2 has removed 4
openssl-dev> symbols compared to 1.0.1, and thus is not strictly ABI compatible,
openssl-dev> linking to <a class="moz-txt-link-freetext" href="https://abi-laboratory.pro/tracker/timeline/openssl/">https://abi-laboratory.pro/tracker/timeline/openssl/</a> . If I
openssl-dev> start semi-randomly clicking around, I can find a page [1] that seems
openssl-dev> to claim the missing symbols are:
openssl-dev> ASN1_STRING_clear_free()
openssl-dev> ENGINE_load_rsax()
openssl-dev> SRP_user_pwd_free()
openssl-dev> SRP_VBASE_get1_by_user()
I haven't make a complete analysis over versions, just did a
comparison of the files that define what we regard as public symbols
(util/libeay.num and util/libssl.num) in the latest 1.0.1 and 1.0.2
releases. Diffs attached.
As you can see, ENGINE_load_rsax *did* go away. That happened here:
commit 74184b6f21e095dacd6193a78785a47dd515f0dc
Author: Dr. Stephen Henson <a class="moz-txt-link-rfc2396E" href="mailto:steve@openssl.org"><steve@openssl.org></a>
Date: Sun Dec 1 23:06:33 2013 +0000
RSAX no longer compiled.
I'm afraid I can't remember the reasoning behind this commit...
The rest of those mentioned above haven't moved at all as far as I can
see. You may notice that some of the symbols in libssl (ssleay.num)
did move between "modules" (which in this case can be defined as a
keyword you can say no to when configuring). I'm unsure how that
affects your view on our stability, suffice to say that with default
configuration, it doesn't affect the ABI one bit.
</pre>
</blockquote>
<br>
<br>
Agreed. Though, just the presence of a function/symbol does not
preclude ABI changes for that symbol, if the function signature (or
behavior) changed.<br>
<br>
-Ben<br>
</body>
</html>