[AWS] Follow up to "Conditional jump or move depends on uninitialised value(s) when GZip'ing content - bug in AWS or?"

[Thomas Løcke]

On Fri, Apr 1, 2011 at 2:46 PM, Dennison, Ted
<Ted.Dennison at flightsafety.com> wrote:
> Thomas Løcke wrote:
>> I tried compiling AWS with ZLIB=false set in makefile.conf and then
>> the problem completely went away.
> ...
>
>> What can be deduced from that?
>
> Err...zlib has an issue? Perhaps you should be talking with the zlib people instead?


I'd buy that, except the problem is eliminated by reverting AWS to this commit:

    commit 1915ca69cc73e29088cefe7cd615b28b7d5de1ba
    Author: Pascal Obry <obry at adacore.com>
    Date:   Fri Mar 11 14:20:53 2011 +0100

        Update 0208_cache_page_server test to pass on VxWorks.

        Part of J817-008.

And the problem comes back with this commit:

    commit 2120bf7cf5f6c071ee9722a5e1a9184554922466
    Author: Pascal Obry <obry at adacore.com>
    Date:   Fri Mar 11 15:11:35 2011 +0100

        Always copy the static libz.

        This library is always built for AWS own use. We now install it to
        cover the case where the dynamic libz is installed in the OS but
        not the static version. If the OS static version is present it
        will be used anyway.

        Should fix K310-011.

So if this is a straightforward problem with zlib, then why can I make
it go away by using AWS commit
1915ca69cc73e29088cefe7cd615b28b7d5de1ba? Is it perhaps something with
the zlib that is bundled with AWS?

I'm more than willing to accept that this issue is due to zlib or
valgrind, but since it appears to be triggered by a _specific_ AWS
commit, then I thought I'd ask here first.

Also I was wrong in my previous email: The problem does not go away if
you use ZLIB=false in makefile.conf. It just went away for some of the
resources in my AWS server.

:o)
Thomas Løcke