This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: hot/cold partitioning bug fix



On 19/11/2004, at 1:56 PM, Mike Stump wrote:


On Thursday, November 18, 2004, at 02:51 PM, Geoffrey Keating wrote:
Mike Stump <mrs@apple.com> writes:

Here is a hot/cold partitioning bug fix that is extremely safe.  I
didn't author this, but would like to see it get in.

Ok?

This needs a testcase.

That's going to be unfortunate, as I think the testcase was SPEC, and otherwise, I don't happen to have one. I tried to re-create one, but was unable. I'll ask around and see if Dale or Caroline has one.


How did you test the patch?

I didn't, the bug was rediscovered by Dale during internal testing. The patch has been tested internally by Dale and Caroline, but, exactly how I am not sure I can say, other than roughly, SPEC.

You did at least bootstrap and run dejagnu, right?


In particular, did you run the C++ testsuite on Darwin with flag_reorder_blocks_and_partition set?

Nope; C++ and reordering don't get along, as C++ uses EH, and EH turns it off.

Try C++ and -fno-exceptions?


(I don't know if I said this originally or forgot, but what I'm concerned about is that text_section() is not correct for a coalesced function.)

I'm trying to get this in, as there are a host of outstanding hot-cold bug fixes that are not getting in, and are being hit by others outside Apple. I think that shipping hot-cold partitioning in 4.0.0 without the bug fixes that we already know about would be unfortunate, specially when we've already found the bugs and fixed them. This is why I am trying to help out.

I decided to try and help get this one in, as trivially, it cannot impact non-hot-cold compiles, and we already know that it helps with hot-cold compiles. I was hoping that I could get this in without the testcase.

It really really needs a testcase. Clearly there must be no existing tests for it, or they'd be failing. It's important to get the bugs fixed, but it's just as important to have at least some tests or the bugs will just reappear.

Attachment: smime.p7s
Description: S/MIME cryptographic signature


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]