llvm dev - Jan 2019 - llvm-lit output directory - role?

Hi,
I use llvm-lit in my own project. Works fine so far.
However, executing the tests, I get output directories which are empty.
Whats the role of them? I could not find any hint in the docu.

I was expecting that e.g. for RUN: X, the output of X would be located
there.
I can use RUN: X > log, then I get a log file right besides the test
file, and not in oputput.

Where can I find more docu about the output directories?

Thx
Alex

On Mon, 28 Jan 2019, alexp via llvm-dev wrote:
> I use llvm-lit in my own project. Works fine so far.
> However, executing the tests, I get output directories which are empty.
> Whats the role of them? I could not find any hint in the docu.
>
> I was expecting that e.g. for RUN: X, the output of X would be located
> there.
> I can use RUN: X > log, then I get a log file right besides the test
> file, and not in oputput.
The normal pattern (within llvm) to use lit tests is that they shouldn't 
write into the current directory. Instead use %t to get a base for a 
temporary pathname (which resides within the output directory). Here you 
could do e.g. "RUN X > %t.log".
> Where can I find more docu about the output directories?
I'm not aware of any docs specifically about the output directory in 
itself, but 
https://llvm.org/docs/CommandGuide/lit.html#pre-defined-substitutions 
describes the substitutions you can use to refer to source/temp paths.

// Martin

Ok,

now it's clear, thanks!
Alex

On 1/28/19 12:15 PM, Martin Storsjö wrote:
> On Mon, 28 Jan 2019, alexp via llvm-dev wrote:
> 
>> I use llvm-lit in my own project. Works fine so far.
>> However, executing the tests, I get output directories which are empty.
>> Whats the role of them? I could not find any hint in the docu.
>>
>> I was expecting that e.g. for RUN: X, the output of X would be located
>> there.
>> I can use RUN: X > log, then I get a log file right besides the test
>> file, and not in oputput.
> 
> The normal pattern (within llvm) to use lit tests is that they
shouldn't
> write into the current directory. Instead use %t to get a base for a
> temporary pathname (which resides within the output directory). Here you
> could do e.g. "RUN X > %t.log".
> 
>> Where can I find more docu about the output directories?
> 
> I'm not aware of any docs specifically about the output directory in
> itself, but
> https://llvm.org/docs/CommandGuide/lit.html#pre-defined-substitutions
> describes the substitutions you can use to refer to source/temp paths.
> 
> // Martin
>