tlf-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Tlf-devel] TravisCI - testing


From: Thomas Beierlein
Subject: Re: [Tlf-devel] TravisCI - testing
Date: Tue, 30 Jan 2018 14:13:18 +0100

Hi Zoli,

your approach sounds not bad. If we need to go one step further we could

a) build a static library from the object files in src and
b) link against it.

It has the nice point that we can override every function from the
library with our own stubs (e.g. in your mentioned global.c) without
getting a double definition error from the compiler.

73, de Tom DL1JBE

Am Sun, 28 Jan 2018 20:55:17 +0100 schrieb Csahok Zoltan
<address@hidden>:

> Hi Tom,
> 
> I have checked your approach. I think the following combination of
> both could work
>     - one test harness like my run.c + the (generated, if poss) test
> defs
>     - global data (mostly from src/main.c) in a global.c
>     - functions that use network/hamlib/curses mocked/wrapped in
> global.c
>     - per file test files (test_xxx.c), each with init/clean as needed
>         * there shouldn't be any interference between tests as long
> as they set up everything they need
>         * a file tests the "public" functions defined in the
> corresponding .h file
>     - all used objects are linked from src
>         * no compilation is needed and
>         * they are used as they are present in tlf executable
>     - Makefile.am is used to set the right libs for linking
>         * mainly needed for glib (I didn't need this unil now)
> 
> Will make a prototype in test-travis to see if the above is feasible.
> 
> 73,
> Zoli
> 
> 
> On Sun, Jan 28, 2018 at 04:35:53PM +0100, Thomas Beierlein wrote:
> > Hi Zoli,
> > 
> > I just pushed my own 'tests' branch which I mentioned yesterday to 
> > 
> > https://github.com/dl1jbe/tlf/tree/tests
> > 
> > I uses cmocka-1.1.1 as test framework. It needs to be installed by
> > hand before, as I did not integrate the check for it into
> > configure.ac until now. 
> > 
> > Maybe we can integrate both our test setups.
> > 
> > 
> > By the way I like your idea with the automatic generation of
> > "defs.h" and the Makefile on the fly. I fear it will only work for
> > tests where the code under test is located in one c-file only.
> > There is still a lot of code which is tightly coupled and needs to
> > compile in more than one file and stub out quite some other
> > functions too (see my code).
> > 
> > One of my goals is to reduce that coupling to make testing easier.
> > 
> > 73, de Tom DL1JBE
> > 
> > 
> > Am Fri, 26 Jan 2018 20:31:00 +0100
> > schrieb Csahok Zoltan <address@hidden>:
> >   
> > > Tom, cloned your repo and Travis CI works quite well.
> > > I'll try to add a CUnit test as per docs. Not all of Tlf
> > > can be tested that way, but let's see how far we can get.
> > > 
> > > The .astylerc you shared looks OK to me.
> > > (yes, I meant indent of 4 spaces)
> > > 
> > > 73,
> > > Zoli
> > > 
> > > 
> > > On Fri, Jan 26, 2018 at 02:42:41PM +0100, Thomas Beierlein
> > > wrote:  
> > > > That repo was only set up as a playground to test the features
> > > > of travis-ci - not for real tlf development.
> > > > 
> > > > 73, de Tom    
> > > 
> > > _______________________________________________
> > > Tlf-devel mailing list
> > > address@hidden
> > > https://lists.nongnu.org/mailman/listinfo/tlf-devel  
> > 
> > 
> > 
> > -- 
> > "Do what is needful!"
> > Ursula LeGuin: Earthsea
> > --
> > 
> > 
> > _______________________________________________
> > Tlf-devel mailing list
> > address@hidden
> > https://lists.nongnu.org/mailman/listinfo/tlf-devel  
> 
> _______________________________________________
> Tlf-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/tlf-devel



-- 
"Do what is needful!"
Ursula LeGuin: Earthsea
--




reply via email to

[Prev in Thread] Current Thread [Next in Thread]