Red Hat g++ packaging question

[Peter J. Stieber]

PS = Pete Stieber
PS>> I looked at the Fedora 10 page plans, but I didn't see
PS>> any updating plans.  Is there any way to see if Fedora
PS>> 10 is planning a GCC 4.3.3 update?

RS = Rahul Sundaram

RS> Unlikely. Build tools are usually not updated post-release.

4.3.2 -> 4.3.3 from the GCC web site is a "regression fixes and docs 
only" only.  It is not a new release.  This has happened in prior Fedora 
releases.  This is one of the things I like about Fedora.

Now if I have to wait for Fedora 11 to get GCC 4.4.x the process will 
start all over again, because this is a new release of GCC and it will 
take a few minor versions updates before they are worked out.  With your 
"Build tools are usually not updated post-release" argument there may 
not be a bug fixed / stable version of GCC for some time...  Unless 
4.3.x-compat is released in Fedora 11.

PS>> Does the Red Hat version of GCC recompile a source module if it
PS>> encounters an error on the first try?

RS> No, GCC behaviour doesn't change like that.

According to Jakub it does...

<Jakub>
Yes, if it encounters an internal compiler error, the gcc/g++ compiler 
driver retries 3 times to determine if a bug is reproduceable or not. 
If not, the most likely culprit is faulty hw, if yes, it prepares 
preprocessed source together with options used and tells you to file a 
bug with that data.
</Jakub>

And this is what I experienced with my random errors.

I just got another random error in last night's build.  This one looks 
like corrupt assembly code.

/tmp/ccmigTCg.s: Assembler messages:
/tmp/ccmigTCg.s:9718: Error: unknown pseudo-op: 
`.size^A_znkst6vectoripk22tccommunicationsdevicesais2_ee8max_sizeev'

Maybe this is some type of disk problem.

Pete