dmd test coverage (page 2)

On 11/28/2014 11:35 AM, H. S. Teoh via Digitalmars-d wrote: > On Fri, Nov 28, 2014 at 11:32:25AM -0800, H. S. Teoh via Digitalmars-d wrote: >> On Thu, Nov 27, 2014 at 09:05:16PM -0800, Walter Bright via Digitalmars-d wrote: >>> On 11/27/2014 4:54 PM, H. S. Teoh via Digitalmars-d wrote: >>>> How do I locally test dmd for coverage? I have a fix for the missing >>>> coverage for version.c, but I can't figure out how to test whether >>>> it's actually working. >>> >>> In posix.mak for dmd, there's 'make gcov' to generate coverage data. >> >> I get a whole bunch of errors that say 'cannot open notes file'. What >> does that mean? >> >> Also, don't I have to first compile dmd with --coverage and run the test >> suite before it has the coverage data?? > [...] > > Which I had tried (make -f posix.mak -j6 MODEL=64 CFLAGS=--coverage), > but the compilation aborts with: > > inifile.c:101:3: error: #error SYSCONFDIR not defined > > which goes away when --coverage is not specified. What gives? I don't know. Haven't tried it in a while, likely it is suffering from bit rot.

November 28, 2014

Re: dmd test coverage

Posted by H. S. Teoh
in reply to Walter Bright

Permalink

H. S. Teoh

Posted in reply to Walter Bright

Permalink

On Fri, Nov 28, 2014 at 12:42:06PM -0800, Walter Bright via Digitalmars-d wrote:
> On 11/28/2014 11:35 AM, H. S. Teoh via Digitalmars-d wrote:
> >On Fri, Nov 28, 2014 at 11:32:25AM -0800, H. S. Teoh via Digitalmars-d wrote:
[...]
> >>Also, don't I have to first compile dmd with --coverage and run the test suite before it has the coverage data??
> >[...]
> >
> >Which I had tried (make -f posix.mak -j6 MODEL=64 CFLAGS=--coverage), but the compilation aborts with:
> >
> >	inifile.c:101:3: error: #error SYSCONFDIR not defined
> >
> >which goes away when --coverage is not specified. What gives?
> 
> I don't know. Haven't tried it in a while, likely it is suffering from bit rot.

Nevermind, I found the reason. Make overwrites CFLAGS even if you intended to append to it with `make CFLAGS+=--coverage`, so the other necessary CFLAGS were missing.  Gah.

Eventually I saw that specifying DEBUG=1 and ENABLE_PROFILING=1 causes the makefile to do the Right Thing(tm), so that's what I did. After that, make gcov does its job.  Except it only works if you're in dmd/src, since the top-level makefile has no idea what you're talking about, in spite of the fact that the two makefiles are obviously connected.  I hate make so much[1]. :-(

[1] http://www.conifersystems.com/whitepapers/gnu-make/

T

-- 
People demand freedom of speech to make up for the freedom of thought which they avoid. -- Soren Aabye Kierkegaard (1813-1855)

On Fri, Nov 28, 2014 at 01:55:26PM -0800, H. S. Teoh via Digitalmars-d wrote: [...] > Eventually I saw that specifying DEBUG=1 and ENABLE_PROFILING=1 causes the makefile to do the Right Thing(tm), so that's what I did. After that, make gcov does its job. [...] I've added this information to the wiki so that it won't be forgotten again: http://wiki.dlang.org/Building_DMD#DMD_profiling_and_test_coverage T -- Just because you survived after you did it, doesn't mean it wasn't stupid!

On 11/27/2014 3:32 AM, Martin Nowak wrote: > Actually not too bad :). > > https://dlang.dawg.eu/coverage/ > https://coveralls.io/r/MartinNowak/dmd And quickfur (H.S. Teoh) once again proves how important having coverage testing is! In trying to increase coverage a lurking bug was uncovered and quickly fixed: https://github.com/D-Programming-Language/dmd/pull/4171

On 11/28/2014 2:59 PM, H. S. Teoh via Digitalmars-d wrote: > I've added this information to the wiki so that it won't be forgotten > again: > > http://wiki.dlang.org/Building_DMD#DMD_profiling_and_test_coverage Good!

"Walter Bright" wrote in message news:m5b044$1bh2$1@digitalmars.com... > And quickfur (H.S. Teoh) once again proves how important having coverage testing is! In trying to increase coverage a lurking bug was uncovered and quickly fixed: > > https://github.com/D-Programming-Language/dmd/pull/4171 Not that fixing it wasn't useful, but that was a minor diagnostic bug that only happens when misusing a feature that nobody uses.

On 11/29/2014 4:08 AM, Trass3r wrote: > Plenty of "not covered" lines are actually assert(0)'s. Right, and some of them are also in #ifdef DEBUG / #endif pairs. This is why the code will never reach 100% coverage. But there's still an awful lot we can get covered, and without much effort, either.

On 11/27/2014 3:32 AM, Martin Nowak wrote: > Actually not too bad :). > > https://dlang.dawg.eu/coverage/ > https://coveralls.io/r/MartinNowak/dmd More tests to improve coverage: https://github.com/D-Programming-Language/dmd/pull/4191 when will this page get updated? https://dlang.dawg.eu/coverage/src/lexer.c.gcov.html

Forums