[openssl-dev] 1.0.2-stable broken Windows build?

[Andy Polyakov]

> Thanks for responding.  Which tool chain are you using?  I'm using VS
> 2013 with the ml compiler.  Given this assembly is generated by a perl
> script, maybe it's a perl issue.  Which perl interpreter are you using?

Quoting INSTALL.W32:

"Netwide Assembler, a.k.a. NASM, available from
http://nasm.sourceforge.net/ is required if you intend to utilize
assembler modules. Note that NASM is now the only supported assembler."

You can drop nasm.exe to any place on %PATH% to get it working.

This means that we formally don't accept problem reports for ml. One can
argue that we ought to, because ml might be not as bad as it used to.
But if we change this, it would probably be more appropriate to do in
post-1.0.2 context. But it doesn't harm to ask for more details. I tried
to emit code for ml and lines in question, i.e. 1432 and 1576, appear
totally innocent, as there are similar instructions next to them. How do
these lines look like for you? I mean could you just copy-n-paste say 5
lines fragments around those line numbers?

>>> Given the 1.0.2 release is forthcoming in the near future, it would be
>>> good if someone could look at this issue.  It looks like there were a
>>> lot of changes made to sha1-586.pl in 1.0.2 to support the new Intel SHA
>>> extensions,

SHA-extension-specific code doesn't require explicit support from
assembler, or in other words, it can be assembled with assembler that is
not aware of SHA extensions. This is because those instructions are
encoded "manually" with DB/.byte directive.

>>> which aren't available until Skylake.

For reference, first processor that will have SHA extensions is
Goldmont, Silvermont's successor, i.e. one from Atom family.