Re: Compilation problem with icecat 3.0.2-g1

[Giuseppe Scrivano]

Michael Cree wrote:

> Notably the above "errors" are in fact warnings however the "treat
> warnings as errors" option was set by the Makefile.  I could disable the
> "treat warnings as errors" option in the Makefile, but I presume that it
> has been prescribed for a reason, and that this is a "dangerous" module
> to be entertaining warnings.
> 
> All of the warnings are about increasing the required alignment of
> target types of pointers.  This does represent a general problem for
> certain CPUs such as the Alpha.  Certain data can only be accessed on
> certain memory alignments. Failure to do so is not completely fatal; it
> normally results in a kernel trap and the kernel, via a sequence of CPU
> instructions, emulates the required access, with an enormous loss in
> performance. It is considered good practice to avoid this situation.
> 
> I guess my initial questions are:  Why has the "treat warnings as
> errors" been set for this module?  Can I safely disable it?

I think you can treat warnings normally without any problem, a little
loss on performance it is better than don't compile at all :)

I guess warnings are handled as errors here because they want the
hashtable implementation to be as fast as possible, if you continue to
have the same error then it is a good idea to report it to the Mozilla
Firefox developers and have the problem fixed there.

Do you get the same error without -O2?

Giuseppe