Elements of branch coverage

[Alexandra]

Hello,

I am interested in the use of JaCoCo for spectrum-based fault localization, especially the branch coverage. However, when testing JaCoCo on a small example, I see that the results are given only in terms of the execution counts. 

The XML file provides more information (mi/ci/mb/cb) but only specifies the source of the branches, not the target. E.g.,

<line nr="11" mi="0" ci="5" mb="1" cb="1"/>

In this case, we can see that branch (11,12) is missed and (11,13) is covered, but I assume that JaCoCo must record the specific branches to ensure that all of them are covered. If I could get any advice on where to find the branch elements in JaCoCo, it would be greatly appreciated.

Kind regards,

Alexandra

[Marc Hoffmann]

Hi Alexandra,

JaCoCo works on compiled Java byte code. So internally it knows about the exact branches of the byte code execution flow. As there is no trivial mappingback  to the source code we do not report this information.

Long time ago there was a prototype to reconstruct this information from additional information that the Java compiler can adds to class files (CharacterRangeTable):

https://github.com/jacoco/jacoco/pull/317

Please also note that JaCoCo does not track execution count. It only tracks whether instructions and branches are executed at least once.

Best regards,

-marc

On 18. May 2026, at 17:13, Alexandra <alexand...@gmail.com> wrote:

Hello,

I am interested in the use of JaCoCo for spectrum-based fault localization, especially the branch coverage. However, when testing JaCoCo on a small example, I see that the results are given only in terms of the execution counts. 

<Screenshot from 2026-05-18 17-11-53.png>

The XML file provides more information (mi/ci/mb/cb) but only specifies the source of the branches, not the target. E.g.,

<line nr="11" mi="0" ci="5" mb="1" cb="1"/>

In this case, we can see that branch (11,12) is missed and (11,13) is covered, but I assume that JaCoCo must record the specific branches to ensure that all of them are covered. If I could get any advice on where to find the branch elements in JaCoCo, it would be greatly appreciated.

Kind regards,

Alexandra

--
You received this message because you are subscribed to the Google Groups "JaCoCo and EclEmma Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to jacoco+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/jacoco/0ec9b706-81d5-4a00-856b-1490d4dbfd37n%40googlegroups.com.
<Screenshot from 2026-05-18 17-11-53.png>

[Alexandra]

Hi Marc,

Thanks for the quick reply! That makes sense, I'll have a look at that prototype.

Could I ask what the numbers in mi/ci then correspond to? E.g., in the above and for four test cases, we have mi=0 and ci=5 for line 11. There's also

<line nr="16" mi="0" ci="2" mb="0" cb="0"/>

This is for my own curiosity more than anything, since in the fault localization case, it only matters whether something is covered or not :)

Kind regards,

Alexandra

[Marc Hoffmann]

Hi Alexandra,

regarding the meaning of the numbers, please find documentation here:

• https://www.jacoco.org/jacoco/trunk/doc/counters.html
• https://www.jacoco.org/jacoco/trunk/coverage/report.dtd

So mi/ci is “missed instructions” and “covered instructions” where instructions refers to Java byte code instructions.

Best regards,

-marc

To view this discussion visit https://groups.google.com/d/msgid/jacoco/33e8c4cd-20c4-4932-8a86-8a1bb956253dn%40googlegroups.com.