Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <048801cc7880$d46e09b0$7d4a1d10$@net>
Date: Wed, 21 Sep 2011 12:06:39 -0500
From: "jfoug" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: RE: 1.7.8-jumbo-6-RC* on Mac OS X

>From: Solar Designer [mailto:solar@...nwall.com]
>
>Jim -
>
>It's totally weird.  Turns out that, for example, opt_errors[1] (pointer
>to string "Unknown option") is not valid right at main() startup, unless
>I change declaration from:
>
>static char *opt_errors[] = {
>
>to:
>
>static const char * const opt_errors[] = {
>
>I suspect that some constructor function in some library corrupts our
>initialized static data.
>
>I think I am going to release -jumbo-6 as-is, considering this a problem
>in Mac OS X or in Xcode (specific versions) rather than in John. %-)
>
>What you're observing on Cygwin is probably different.

Likely, but it may be the same 'symptoms', where something at startup is
corrupting certain data.

I do not see the 'non-mmx' cygwin build being any form of show stopper. I
only use it as a 'quick' check for non-sse systems.  Now that I have some
other systems that are non x86, I use that build less and less.  However, it
'used' to work.

I will take a little more time, and static look at code, (especially some of
the newer stuff), and see if there is somewhere that we simply use too much
stack space, somewhere.  I may also try to add some stack size options
(-Wl,stack,size or something like that).

Jim.

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.