Michael Van Biesbrouck wrote:
On 8/23/06, Steve Reinhardt <[EMAIL PROTECTED]> wrote:
Michael Van Biesbrouck wrote:
> On 8/23/06, Steve Reinhardt <[EMAIL PROTECTED]> wrote:
>> Yea, I just noticed that in spite of the name, --enable-shared
seems to
>> create a non-shared version of the library as well. (Or maybe the
>> non-shared version always gets built?)
>
> The non-shared library is always built, but it is never usable for
> embedded purposes (thus the errors that were being reported).
Did you see that documented somewhere, or are you just saying that based
on your experience? I saw this note here:
http://www.python.org/doc/ext/link-reqs.html
that sounded like it might explain the problems you were having with the
static library, but didn't actually try it to see if it helped.
That might fix the problems that I have experienced. The purpose
there is to support loading shared objects for dynamic extensions, so
I'm not sure that any benefit would occur (dynamic linking will still
occur even though the static library was used). It might remove the
LD_LIBRARY_PATH requirement that I have.
Right, it may fix the short-term issue but it doesn't fully address the
general problem of building a single self-contained binary. I believe
the answer there is that you need to build a custom libpython*.a that
statically links in all of the Python library modules you intend to
import as well. According to the README in the Python distribution,
this can be done by editing the Modules/Setup file in the source tree
before you build. If you decide to try this, please let us know how it
goes.
Is the counter initialization bug easy to fix? I would like to start
using M5 as soon as possible.
It's an easy fix (already done in our tree); the bigger effort is
packaging up a new release that includes that and the few other patches
we've had to make to 2.0b1. Those bogus stats are the only thing the
patch fixes, so unless those particular statistics are crucial to what
you're doing, there's no reason not to go ahead and use 2.0b1.
FYI, we're still working on putting up a public Mercurial repository to
make it easier for us (or others!) to push & pull patches without having
to go through an explicit release process.
Steve
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
