vtable thunks (was: RE: building egcs under LinuxPPC pre-R5)

[Franz Sirl <Franz.Sirl-kernel@lauterbach.com>]

At 10:40 10.02.99 , Gary Thomas wrote:
>The latest snapshots are very unstable on Linux/PPC.  We are working on
>this now, but I'd stay away from them until further notice.
>
>On 10-Feb-99 Andreas Tobler wrote:
>> Hello,
>> 
>> since the beginning of February I try to build egcs snapshot 19990131
>> and now 19990208. The succcess is small. I got the 19990131 snapshot to
>> build and install, but when I compile c++ code with some include
>> restrictions it seems that it overrides it which causes an error. The
>> egcs I used before and now again, was 2.91.60 I think, for LinuxPPC
>> folks egcs-1.1.1.1c. With this compiler I haven't had this error.
>> My question, is there something specific when building an egcs on
>> LinuxPPC that I have to take care of? It's not necessary for me to take
>> a snapshot, but I need to build an egcs & libstdc++ with the option
>> -fno-vtable-thunks.
>
>The version of EGCS-1.1.1c that I uploaded should have been compiled this
>way.  What makes you think it was not?

He's probably talking about the pre-R5 egcs-1.1.1-1c RPM's which I prepared
for the R5 distribution. They have vtable-thunks enabled, as it is the
default for glibc-2.1 based egcs.
Gary, if you are producing RPM's for R5 which have vtable-thunks disabled,
I hope you are using versioning to distinguish the binary incompatible
libstdc++ stuff.

I would also appreciate if someone knowledgeable would document
(history/future, pros/cons, etc.) the whole vtable-thunks/no-vtable-thunks
stuff and put it in the FAQ.html (or a separate vtable.html?), cause
searching the mailinglists reveals 140 :-) documents on this issue.

Franz.