[LLVMdev] !!! 3.2 Release branch patching and the Code Owners
wendling at apple.com
Fri Nov 16 14:56:03 PST 2012
On Nov 16, 2012, at 2:17 PM, 32bitmicro <root at 32bitmicro.com> wrote:
> Recent code owner activities have led to
> what I would call loss of referential integrity
> in the CODE_OWNERS.TXT file.
> Changes are fine but the information in the
> CODE_OWNERS.TXT does not allow to positively
> identify code owner of the particular
> file or patch.
> The problem stems from the usage of the
> "description (D)" field which is overloaded
> with meaning. Most people put only a textual
> description of the code they own.
> This approach is fine for casual reader but
> does not work for scripting or any automated
> way of dealing with the build.
> I would like to propose addition of the
> "folder/file (F)" field. The format
> would be the same as used by Joe,Owen
> and Justin
> F: (path/relative/to/llvm/src/root/*)
> F: (path/relative/to/llvm/src/root/file.name)
> F: (lib/Bitcode/* include/llvm/Bitcode/*)
> F: (lib/CodeGen/SelectionDAG/*)
> F: (lib/Target/NVPTX/*)
> Situation is particularly bad for the 3.2 branch as
> "old" owner is not necessarily the same as new one.
> So for the time being I have adopted the policy
> of using code owner from the trunk as the one
> responsible for approving the patches.
> This worked for a while but there are more and
> more patches requested with no clear way of
> identifying the owner. Situation has got to
> the point where AI (lame as it is) embedded
> in the 3.2 integration bots simply says:
> DOES NOT COMPUTE!
> Therefor I have no choice but to suspend
> accepting all of the 3.2 patches until the
> situation gets resolved.
That's a bit draconic. The code owners file is a text file and not important to the execution of LLVM. Therefore, it's fine to omit patches for it if there are conflicts.
More information about the llvm-dev