Glenn A. Thompson wrote:
> Hey,
>
>>
>> Anyway, HP-UX is very much alive, is being ported to IA64 (in fact, it
>> already runs on that platform), so it's got a few years of life left
>> yet. Given that this problem occurs on the 11.x series, too (not just
>> 10.20), I believe it has to be fixed.
>
> The FAQ on sleepycat implies that it is a hardware problem. It seems
> hard to believe that a HP-UX patch couldn't be developed to fix the
> problem. If it is a hardware problem then maybe it will work on IA64.
Indeed, so it does. Although I can't figure out why another map of the
same file wouldn't just return the pointer to the existing map... duh.
>> The fact that people haven't been
>> pushing for the fix is probably just because mose people who'd use
>> Subversion right now are on *BSD or Linux or Mac OS. Even Solaris isn't
>> that well represented, HP-UX and AIX are definitely stage left.
>>
>>
> I do think the fix should go in. I just disagree with the approach
> you outlined.
> As for Solaris. I, for one, am a *huge* Solaris fan. I run Solaris
> on Intel and Sparc. For current development purposes, I feel it is
> easier for me to develop on Linux. Any production repos I put in
> *will* be Solaris.
I only meant to say that these systems aren't as well represented now,
not that they won't be represented after 1.0.
>> As for the "just one platform" argument: When the fix goes in (and it
>> will, depend on it), it must either work on *all* platforms, or be
>> conditionalized. If people think we can live with conditional code for
>> the fix, then I guess we can also live with using the abspath of the FS
>> as the key, even though that's not completely unique.
>>
>>
> I disagree with the "or" above. The fix can work on all platforms
> *and* be conditionalized.
Well, yes, it would work on all platforms, but we might not want to use
it unless its necessary. Depends on the fix, of course; I've given it a
bit of thought, but not nearly enough to start working on it. The pool
lifetime problem sounds interesting, but solvealbe (no, not soluble :-).
> Sorry to be such a pest, but I'm *sure* you won't hold back on me
> when the opportunity presents itself. :-)
Not to worry. How does it go already? "A friend with weed is a friend
indeed, but a friend in need is a pest!" I'm quite happy to cater to
your needs. :-p
--
Brane Čibej <brane_at_xbc.nu> http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 18 20:41:07 2003