Joerg Schilling wrote:
> Wyllys Ingersoll <Wyllys.Ingersoll at Sun.COM> wrote:
>
>> Wouldn't the logical replacement for Pth be libpthread or Solaris threads?
>> Why bother even attempting to write a new thread library when we have
>> viable alternatives? The goal should be to eliminate the need for Pth
>> in the programs that use it if possible.
>
> Please read the other mails from this thread. The Pth library does not
> implement
> cuncurrency, so it cannot be replaced by libpthread without running the risk
> of
> causing problems in the software that was implemented to call Pth.
>
Before this PSARC case I knew nothing about GNU Pth. I made what I
thought were reasonable assumptions about the semantics of this library,
but after comments like the one above (and a little online research), I
agree that it could be unsafe to implement it as a libpthread wrapper.
Never mind.
To answer Wyllys's question, the rationale for providing a Pth
implementation (including a more efficient one if it were safe to do so)
would be to provide it for other programs to link against. Providing the
expected set of system libraries is arguably the most important form of
"Linux familiarity."
That motivation notwithstanding, modifying GnuPG to use libpthread is a
fine idea.
Scott
--
Scott Rotondo
Principal Engineer, Solaris Security Technologies
President, Trusted Computing Group
Phone/FAX: +1 408 850 3655 (Internal x68278)
