This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Workaround for bugs 42614 and 42691 in GCC 4.4.2
- From: Jeff Law <law at redhat dot com>
- To: Vicent Brocal Tortosa <vbrocal at fentiss dot com>, gcc at gcc dot gnu dot org
- Date: Thu, 16 Aug 2018 09:50:37 -0600
- Subject: Re: Workaround for bugs 42614 and 42691 in GCC 4.4.2
- References: <7E7329D4-10D2-429E-9D71-CB81A69B0846@fentiss.com>
On 08/16/2018 09:43 AM, Vicent Brocal Tortosa wrote:
> Hello,
>
> We are forced to use version 4.4.2 of GCC C compiler for a certain project and to analyze the potential impact, to our software, of bugs detected for this GCC version.
>
> I am already aware that bugs reported for 4.4.x versions may also be present in 4.4.2, but for the moment we have decided to keep the scope of the analysis for the ones reported specifically for this version ("know to fail").
>
> In the GCC bugzilla tracker I have found a couple of them that bother me (we intend to use -O2 optimizations):
>
> Bug 42614 - [4.4 Regression] FRE optimizes away valid code after IPA inlining.
> Bug 42691 - [4.4/4.5 regression] problematic REG_EQUAL note added to SUBREG.
>
> Can someone recommend a general workaround for them (e.g. deactivating the affected optimizations) or any other approach that can be used to ensure that there is no impact to the object code?
I doubt there's an acceptable general workaround. THe first would
probably require disabling FRE or IPA which is a pretty big hammer.
THe second you might be able to disable RTL forward propagation.
You're likely better off upgrading or applying the referenced fixes to
your compiler.
gcc-4.4 is horribly old at this point. You're unlikely to get many
resources to help you if things go wrong.
jeff