Peter Eisentraut <pe...@eisentraut.org> writes:
> Here is a patch set to remove what turns out to be a significant amount 
> of code and documentation to support this compiler.

Looks generally sane by eyeball -- I did not grep for anything that
you missed.  I concur with Andres that tas.o is probably never going
to be used again, and if it is we can put back support for it.

> In src/include/storage/s_lock.h, there was a comment to look in 
> src/backend/port/tas/sunstudio_sparc.s for some explanation.  But the 
> latter is to be removed, so I copied the comment over and tried to fit 
> it in.  But I don't understand it much, and it makes some claims about 
> gcc support for sparc, so it could be that some more code for newer 
> sparc variants is dead, or maybe gcc has caught up in the meantime. 

I would reduce the comment in s_lock.h to something along the lines of

    /*
     * "cas" would be better than "ldstub", but it is only present on
     * sparcv8plus and later, while some platforms still support sparcv7
     * or sparcv8. Also, "cas" requires that the system be running in
     * TSO mode.
     */

There's a comment a few lines above explaining TSO, so we don't need
more than that here.

Possibly at some point somebody will be motivated to improve the
s_lock.h code beyond its current state, but I agree that that's
material for future work.

                        regards, tom lane


Reply via email to