Regression testing with lit on a constrained toolchain

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Regression testing with lit on a constrained toolchain

Bakhvalov, Denis via cfe-dev

Forgive me if this is common knowledge amongst the more experienced here, but I’m currently confused by the lit regression tests.

I’m currently working to validate a Clang/LLVM toolchain, hosted on a Mac OS, which only supports a select number of baremetal ARM targets.

 

Specifically, the test Driver/darwin-ld-dedup.c fails due to not having available targets compatible with ‘x86_64-apple-macosx10.6.0’. As far as I’m concerned, this target is a do-not-care for me.

Another that seems suspect is Index/pch-from-libclang.c, which also sets the triple directly to x86_64-apple-darwin.

 

My target triple does indeed look correct (arch is arm), and the host triple is x86_64-apple-darwin*.

 

Lit supports REQUIRED checking the target triple, but such checks just don’t appear in these failing tests.

Is this not a common enough problem, or might I be missing something fundamental in my understanding?

 

Thanks,

J.B. Nagurne

Code Generation

Texas Instruments


_______________________________________________
cfe-dev mailing list
[hidden email]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
Reply | Threaded
Open this post in threaded view
|

Re: Regression testing with lit on a constrained toolchain

Bakhvalov, Denis via cfe-dev
On 02/18/2019 05:13 PM, Nagurne, James via cfe-dev wrote:

> Forgive me if this is common knowledge amongst the more experienced here, but I’m currently confused by the lit regression tests.
>
> I’m currently working to validate a Clang/LLVM toolchain, hosted on a Mac OS, which only supports a select number of baremetal ARM targets.
>
>  
>
> Specifically, the test Driver/darwin-ld-dedup.c fails due to not having available targets compatible with ‘x86_64-apple-macosx10.6.0’. As far as I’m concerned, this target is a do-not-care for me.
>
> Another that seems suspect is Index/pch-from-libclang.c, which also sets the triple directly to x86_64-apple-darwin.
>
>  
>
> My target triple does indeed look correct (arch is arm), and the host triple is x86_64-apple-darwin*.
>
>  
>
> Lit supports REQUIRED checking the target triple, but such checks just don’t appear in these failing tests.
>
> Is this not a common enough problem, or might I be missing something fundamental in my understanding?
>

It's a bug, those tests should also have 'x86-registered-target' added
to their REQUIRES.

-Tom

>  
> Thanks,
>
> J.B. Nagurne
>
> Code Generation
>
> Texas Instruments
>
>
>
> _______________________________________________
> cfe-dev mailing list
> [hidden email]
> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>

_______________________________________________
cfe-dev mailing list
[hidden email]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
Reply | Threaded
Open this post in threaded view
|

Re: [EXTERNAL] Re: Regression testing with lit on a constrained toolchain

Bakhvalov, Denis via cfe-dev

I'll add them locally so I can get up and running, and add it to the queue of things to upstream when we're stable. I'm glad I wasn't missing anything. I suppose a majority of bots just build out-of-the-box and wouldn't catch something like this.


Thanks much,

J.B. Nagurne

Code Generation

Texas Instruments



From: Tom Stellard <[hidden email]>
Sent: Monday, February 18, 2019 9:00 PM
To: Nagurne, James; [hidden email]
Subject: [EXTERNAL] Re: [cfe-dev] Regression testing with lit on a constrained toolchain
 
On 02/18/2019 05:13 PM, Nagurne, James via cfe-dev wrote:
> Forgive me if this is common knowledge amongst the more experienced here, but I’m currently confused by the lit regression tests.
>
> I’m currently working to validate a Clang/LLVM toolchain, hosted on a Mac OS, which only supports a select number of baremetal ARM targets.
>

>
> Specifically, the test Driver/darwin-ld-dedup.c fails due to not having available targets compatible with ‘x86_64-apple-macosx10.6.0’. As far as I’m concerned, this target is a do-not-care for me.
>
> Another that seems suspect is Index/pch-from-libclang.c, which also sets the triple directly to x86_64-apple-darwin.
>

>
> My target triple does indeed look correct (arch is arm), and the host triple is x86_64-apple-darwin*.
>

>
> Lit supports REQUIRED checking the target triple, but such checks just don’t appear in these failing tests.
>
> Is this not a common enough problem, or might I be missing something fundamental in my understanding?
>

It's a bug, those tests should also have 'x86-registered-target' added
to their REQUIRES.

-Tom


> Thanks,
>
> J.B. Nagurne
>
> Code Generation
>
> Texas Instruments
>
>
>
> _______________________________________________
> cfe-dev mailing list
> [hidden email]
> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>


_______________________________________________
cfe-dev mailing list
[hidden email]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev