[Lldb-commits] [PATCH] D61952: [CMake] Stabilize install process for LLDB.framework

Fri May 17 07:24:50 PDT 2019

sgraenitz added a comment.

>> In D61952#1503551 <https://reviews.llvm.org/D61952#1503551>, @JDevlieghere wrote:
>> 
>>> How does this cleanup affect dependency tracking? Does the build dir become unusable after running ninja install?
>> 
>> 
>> In D61952#1504942 <https://reviews.llvm.org/D61952#1504942>, @sgraenitz wrote:
>>  Actually, why not make the copy operations `PRE_BUILD` actions of the test suite instead of `POST_BUILD` actions of their executables?
> 
> In D61952#1505019 <https://reviews.llvm.org/D61952#1505019>, @sgraenitz wrote:
>  The solution to the "install -> test" issue would be having both I guess.

TBH, I don't know how to accomplish this in the current state of the build system. I went through various options and didn't find a functioning one that works without, basically, turning everything upside down. We have similar functionality in `lldb-framework-headers`, but the appraoch only works in the directory where `lldb-framework` was defined. Furthermore we cannot rely on the existence of `lldb-framework` as `lldb_add_to_framework()` may be called before `/lldb/source/API` is processed (e.g. everything else in `/lldb/source`).

On one hand, the main use-case works: if the test suite succeeds then run install. And it sounds "kind of" acceptable that it fails the other way around.
On the other hand, seeing the amount of extra effort, "workaround" may be a better term to describe this change then "stabilize".

At the moment I agree, that neither of the options look really appealing. Good point to rethink the approach and find something more solid. If that fails, I might re-evaluate this one.

@xiaobai Out of interest: have you faced overwrite issues when running `install` and would this patch help?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D61952/new/

https://reviews.llvm.org/D61952