Bug #11762
open
GCC 5 and 6 need -Wno-maybe-uninitialized too
0%
Description
We (RackTop) build illumos with GCC 5.3 and OI Hipster seems to provide GCC 6.X as part of it's build-essential package, but neither of these compilers are able to complete a build because of GCC warning about variables that may be uninitialised.
Updated by Joshua M. Clulow almost 2 years ago
I think we're generally trying to focus on building with the specific GCC branches we're maintaining on GitHub, as outlined in IPD 7 illumos GCC maintenance
I don't think we've ever supported building with GCC 5.X or 6.X -- we moved directly from our supported 4.4.4 to 7.3.0/7.4.0.
If you're building with GCC 5.3 today, which GCC patches/branch are you using to do that?
Are you aware of anybody building with GCC 6.X in practice? If not, I don't think we'd want to encourage it at this stage.
Updated by Andrew Stormont almost 2 years ago
We ported the patches for GCC 4.4.4 over to 4.5, 4.7, 4.9, 5.1 and then 5.3 where we've decided to stay a while, though we'll probably upgrade to 7.X at some point in the near future. I'm not aware of anyone using GCC 6 or any other version specifically. With a minor change to the Makefiles I was able to build illumos-gate using the version of GCC 6 which Hipster ships.
Updated by John Levon almost 2 years ago
I would strongly recommend a prioritisation of a switch to GCC7. Simply put we have zero idea what kind of code untested versions of GCC might be generating, and you could end up with (for example) a serious security issue due to this...
I understand that it's been working for you so far, but I can speak from recent experience things get tricky quickly.
(I would be -1 on anything that tried to support such versions of GCC in illumos-gate itself for these reasons.)
Updated by Garrett D'Amore almost 2 years ago
John Levon wrote:
I would strongly recommend a prioritisation of a switch to GCC7. Simply put we have zero idea what kind of code untested versions of GCC might be generating, and you could end up with (for example) a serious security issue due to this...
I understand that it's been working for you so far, but I can speak from recent experience things get tricky quickly.
(I would be -1 on anything that tried to support such versions of GCC in illumos-gate itself for these reasons.)
I tend to agree .. we are indeed prioritizing a switch to GCC 7, but we have to do it on the next release cycle of our product (stability reasons).
John Levon wrote:
I would strongly recommend a prioritisation of a switch to GCC7. Simply put we have zero idea what kind of code untested versions of GCC might be generating, and you could end up with (for example) a serious security issue due to this...
I understand that it's been working for you so far, but I can speak from recent experience things get tricky quickly.
(I would be -1 on anything that tried to support such versions of GCC in illumos-gate itself for these reasons.)