[PATCH] D120305: [Driver] Default CLANG_DEFAULT_PIE_ON_LINUX to ON

Sat Feb 26 01:08:51 PST 2022

MaskRay added a comment.

In D120305#3347150 <https://reviews.llvm.org/D120305#3347150>, @tstellar wrote:

> [...]
>
>>> It's not acceptable, because buildbots are meant to represent a specific configuration that the buildbot owner cares about.  Changing the buildbot configuration makes the buildbot no longer useful to them since it is now testing something different.  For example, if someone is running production builds that used the same configuration as the buildbot, those production builds are still broken even though the buildbot is green.  Configuration changes like this need to have approval of the bot owner.
>>
>> If clang-ppc64le-rhel works with `-DCLANG_DEFAULT_PIE_ON_LINUX=off` but not with  `-DCLANG_DEFAULT_PIE_ON_LINUX=on`, adding  `-DCLANG_DEFAULT_PIE_ON_LINUX=off` makes the intention explicit. I am not sure why this isn't acceptable.
>> It's a tech debt, though, as we are making configurations fragmented.
>>
>>> I still don't understand why you can't revert the patch.  I've encountered this same situation numerous times while working on this project and no one has ever objected this much to doing a revert.  The fact that this is the second time this has happened is concerning to me.
>>
>> I have stated my reasoning. Configuration churn can also be a problem to users.
>> Consider what if the DWARF v5 patch got reverted and relanded back and forth. Downstream users would keep observing changing behaviors.
>>
>> In this case, really even normal ppc64le machines (including some bots) were happy and just that one was picky.
>> Given that the bot does not have a high success rate (track record) for at least the past month, I am unsure I am supposed to revert my change.
>>
>> It is more concerning to me that a bot maintainer leaves an unstable bot for so long and is not even willing to spend very little time to make a llvm-zorg change live (perhaps just restart the bot software), but rather is more willing to **write such a long reply taking it very personally**.
>> (Sorry, I know when I write this, I was a bit in a mood. I felt quite frustrated at this point, so I could not keep using the tune when I made the reply https://reviews.llvm.org/D120305#3347094)
>>
>> I can response to you that I have shared the thread to some folks and at least two agree with me that nemanjai's reply made the discussion less technical but more personal.
>
> Can these folks provide their feedback on the thread?

Folks willing to provide feedback usually want to be anonymous. You may understand that they don't want to be bothered by further messages and don't want to stand by one side.

I know my replies are wasting subscribers' time and I feel sorry about that.
But I can assure to you that if without https://reviews.llvm.org/D120305#3347058 and your reply to that comment, we were likely in a green state pretty early 🥳

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D120305/new/

https://reviews.llvm.org/D120305