[llvm-dev] bot clang-x86-ninja-win7 missing llvm/lib/DebugInfo/MSF/LLVMBuild.txt

[Hans Wennborg via llvm-dev]

On Mon, Aug 1, 2016 at 9:44 AM, Reid Kleckner <rnk at google.com> wrote:
> I think this is the sequence of events:
>
> - check in directory named Msf
> - use git to rename directory to MSF, all good
> - git doesn't track the existence of directories, but svn does, so git-svn
> dcommit creates a new directory called MSF next to the old Msf directory
> - this all works on case insensitive filesystems
> - Hans tries to clean up the old name manually in svn with 'svn rm'
> - this deletes the 'MSF' directory on case insensitive filesystems (current
> state)
>
> Now 'svn status' is reporting a conflict on these slaves. I'm surprised
> buildbot doesn't automatically recover from conflicts by blasting the
> conflicted portion of the checkout. Maybe we need to reconfigure it to do
> that?

fwiv, the bots were in conflict before I did "svn rm". I suspect the
moment they tried to update to the revision that had both a "MSF" and
"Msf" dir, they were doomed.

(For Chromium bots we tracked this in http://crbug.com/632912)

> On Mon, Aug 1, 2016 at 9:15 AM, Greg Bedwell <gregbedwell at gmail.com> wrote:
>>
>> How about going through an intermediate name?  E.g.   Msf -> foobar -> MSF
>>
>> It's not exactly elegant, but may be easier than fixing all of the bots
>> (presumably this might also affect downstream non-public bots?)
>>
>> On 1 August 2016 at 17:12, Zachary Turner via llvm-dev
>> <llvm-dev at lists.llvm.org> wrote:
>>>
>>> The latter. Unfortunately we need to get physical access and delete the
>>> checkout. I'm afraid reverting will just reintroduce the problem and
>>> re-break all the bots that people already took the time to fix, since going
>>> from MSF -> Msf is the same as going from Msf -> MSF
>>>
>>>
>>> On Mon, Aug 1, 2016 at 8:55 AM Reid Kleckner <rnk at google.com> wrote:
>>>>
>>>> So, the MSF rename managed to break every Windows bot over the weekend.
>>>> Did we try reverting, or was this one of those wonderful situations where
>>>> it's basically the fault of the version control system and reverting doesn't
>>>> help?
>>>>
>>>> On Sat, Jul 30, 2016 at 5:49 PM, Robinson, Paul via llvm-dev
>>>> <llvm-dev at lists.llvm.org> wrote:
>>>>>
>>>>> The bot has been failing its CMake step for a while now.
>>>>> Probably fallout from renaming Msf to MSF.  Can somebody
>>>>> get this straightened out?  It's one of the bots that run
>>>>> VS2013 AFAICT which makes it useful to me.
>>>>>
>>>>> Thanks,
>>>>> --paulr