[Swan-dev] odd error only on ppc64 on rhel6

Paul Wouters paul at nohats.ca
Thu Jun 13 17:14:08 UTC 2019


On Thu, 13 Jun 2019, D. Hugh Redelmeier wrote:

> | > | I'd rather see this case fixed properly by adding the missing
> | > | initializer, so we don't have to have custom flags passed between
> | > | different Linux distributions.
> | >
> | > As I explained last, this is not a bug in Libreswan.  It is
> | > intentional and correct.  I said this last time.
> |
> | Is there a way to rewrite the code not to trigger this issue. The
> | advantage is that we than don't need to track special gcc option
> | passing to various builds we do.
>
> I tried to explain twice: I expect (but haven't tested) that this flag can
> be passed to GCC on every build we do.  Not just old ones.

I missed that option. I thought it would still be useful to not ignore
all of these.

> There's a lot of code that would need to be rewritten to avoid this
> warning.  And it would make the code worse.

No. There is only 1 single instance of this causing a compile failure.

> | > The GCC project thinks this warning is too aggressive.  That's why it
> | > doesn't show up in later versions of GCC.
> | >
> | > This is NOT a custom flag, as far as I know.  Try it.  I bet it works
> | > with all versions of GCC that we care about.
> |
> | I meant custom as "I need to do something custom for RHEL6/CentOS6 builds"
>
> I meant: "you can use this flag universally and don't need to do something
> custom for RHEL6/CentOS6 builds".

That is something we can do, but we would be adding this option on all
our builds to fix 1 specific initializer on one specific platform/gcc
version. That seems excessive and prone to a future "cleanup". But I'm
fine with this solution if that is the best we can do now.

Paul


More information about the Swan-dev mailing list