Please help me : libstdc++-v3 doesn't compile with GCC-2.95.2 on Sparc-Solaris 2.6

Phil Edwards pedwards@disaster.jaj.com
Fri Jan 19 01:24:00 GMT 2001


On Fri, Jan 19, 2001 at 09:36:15AM +0100, Maurizio Loreti wrote:
> On Fri, 19 Jan 2001, Phil Edwards wrote:
> 
> > I assume you meant to type 2001 there.  In any case, we hope to have it
> > finished within the next three months.
> 
> No, I mean exactly 2000.

Yikes.  Are there any web pages or documents on the GCC site which still
say 2000?  If so, that needs to be fixed; please let us know.


> > There was a bug introduced before the 15th of this month.  The snapshot
> > created on the 15th contained that bug.  The bug has since been fixed.
> 
> What do you mean here?  "Download a CVS snapshot; make bootstrap; make
> check; if (errors) wait [unspecified period of time], reload and repeat"?

That's what some of the developers do.  :-)

More seriously:  it depends on what you need, versus what risk is acceptable.

If you need a completely stable and unchanging environment, then you need
to use GCC 2.95.2 and the library that went with it, or any of the new
library snapshots up though (IIRC) 2.90.7.

If you need a more standard-conforming environment and are willing to put
up with some oddities, then use some snapshots.

If you want the *most* bug-free environment and are willing to put up
with occasional instabilities resulting from seriously fatal bugs (which
sometimes take more than a week to resolve), then just use the current CVS.
Definitely subscribe to the mailing lists in this case.

There are lots of possibilities.  (Including using the sources you have
right now and configuring for i586.)


> > Since libstdc++-v3 is part of the g++ sources now, you might try just
> > using current CVS (or a snapshot), and using the library sources that
> > come with it.  They will be more recent than the 2.90.8 library snapshot.
> 
> I assume you misspelled "2.91".

Yes, I did.  :-)


> > Simply configure with --enable-libstdcxx-v3, and you will get the new
> > library.  Also, flags like -fhonor-std will be on by default, so you don't
> > need to set them yourself.
> 
> Well, what about modifying you web page?  Say something as:

This is a good idea.  Thanks for that suggestion.

Rather than specifying the valid snapshots down to the minute of their
creation, I'll probably add a link to the most recent working snapshot.


> If I follow YOUR instructions and I do not get what you promises, don't
> blame ME for having used a buggy snapshot.

There's no blame intended.  But understand that GCC snapshots (unlike
libstdc++-v3 snapshots) are not tested in any way.  They are created
automatically every seven days, with zero human involvement.

Usually the library works with any given GCC snapshot.  The most recent
snapshot (on the 15th) had a serious bug.  The previous snapshot probably
was fine.  The bug was fixed on the 16th, so the next snapshot that will
be created on the 21st should also be fine.

Speaking of snapshots... after Gaby finishes converting the testsuite
to use DejaGNU, maybe we should roll a 2.92 library snap.  It's been two
months since 2.91.


-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.


More information about the Libstdc++ mailing list