Marc Glisse wrote:
On Sun, 17 Jan 2010, Dr. David Kirkby wrote:
Marc Glisse wrote:
It looks like it was fixed in trunk in Septembre (r151331), did you
check?
No I did not. Having seen your note, I tried to download that specific
tag, but did not get anywhere. I have however just downloaded the
latest weekly 4.4 series snapshop
That patch is discussed there:
http://gcc.gnu.org/ml/gcc-patches/2009-09/msg00114.html
It doesn't look like it was backported to the 4.4 branch, but you could
try doing it yourself.
gcc-4.4-20100112.tar.bz2
That was last modified 01/12/10 (I guess the name give you that, but
in a bit of a non-obvious way).
It is much more obvious than the american nonsensical way of writing dates.
I thought the day-month-year was the most logical, increasing in size from left
to right. I agree the American way is silly.
I was not just alluding to the fact this is a date. I think there should be
hypens between the numbers. i.e.
gcc-4.4-2010-01-12.tar.bz2
would have been a bit better. Better still would have been
gcc-4.4-12-01-2010.tar.bz2
The fact the first number starts with a '2', could also lead someone to believe
it was a 4.4.2 revision.
You could also believe this is a patch number, and not a date. Again, it is not
obvious that 20100112 is not a patch number, or something like that.
Or more to the point, what is the best way for me to get this fix,
with the least risk of having a broken compiler?
Wait for 4.5.
That is not really an option. In any case, the 4.x.0's often seem to have major
bugs.
The latest 4.5 snapshot will not build on Solaris, as it uses some options to
'find' which the Sun version of 'find' does not accept. I doubt the options
chosen are POSIX. It does however have the fix integrated. I might be able to
sort out how I can backport it myself.
If not, I intend contacting the original author. I am willing to make a small
contribution to get this fixed. It is quite important for the Sage project.
Some bugs are just a bit of an annoyance, (like the memset bug was on Solaris
with sun4v processors). This is *much* more serious, as it stops the build
process completely, in a place which it is critical (the main library). Had this
been in one of the less important bits, then I would be less concerned.
GCC for Sun systems might be of help, though that is only for SPARC. But that is
somewhat more critical now, as the Sage will run on Solaris 10 SPARC, whereas
the port to x86 is not complete.
Dave
