You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Ginkgo's output can be a bit noisy for large test suites compared to RSpec's -fd output. The ginkgo-fd post-processor is an attempt to achieves this by consuming Ginkgo's --json-report output.
Should I create a PR that adds the -fd flag directly into Ginko?
$ go run github.com/onsi/ginkgo/v2/ginkgo -fd ./reporters/...
DefaultReporter
Rendering SuiteWillBegin
Default Behavior
With Labels
With long Labels
With SemVerConstraints
With ComponentSemVerConstraints
When configured to randomize all specs
when configured to run in parallel
when succinct and in series
when succinct and in parallel
when succinct and with labels
when succinct and with SemVerConstraints
when succinct and with ComponentSemVerConstraints
•••
JunitReport
default behavior
generates a Junit report and writes it to disk
when configured to omit all the omittables
generates a Junit report and writes it to disk
when configured to write the report inside a folder
creates the folder and the report file
Ran 127 of 127 Specs in 0.078 seconds
SUCCESS! -- 127 Passed | 0 Failed | 0 Pending | 0 Skipped
PASS
Ginkgo ran 1 suite in 2.338581833s
Test Suite Passed
If Ginko responds to multiple flags, the current -fd behavior may bypasses it.
• How should the -fd behave with report.RunningInParallel which is currently ignored?
• Should the logic of DidRun() be broken out into new functions or should didRunFd() be embedded into DidRun()?
I've closed PR1669 as PR1670 seems so much better to me.
@woodie I think it's fine having this bypass other flags and essentially be a stand-alone opt-in mode. it occurs to me that it is unlikely to play well with -randomize-all and parallel specs as in both cases there is no guarantee that tests at different container nestings won't interleave.
+-fd (RSpec-style "format documentation" output) is now exclusive of -p/-procs and -randomize-all. Its hierarchical tree assumes specs run serially in declaration order, so Ginkgo now overrides those flags and prints a notice when it does, rather than producing fragmented, hard-to-read output.
Activity
onsi commentedon Jun 4, 2026
sure thing! go for it! be sure to update the docs as well, thanks!
woodie commentedon Jun 4, 2026
We could use the same footer.
vs this
onsi commentedon Jun 4, 2026
imo keep the same footer
woodie commentedon Jun 4, 2026
Will do. A direct clone seemed like a good starting point, but more (unnecessary) work to maintain.
woodie commentedon Jun 5, 2026
OK, the latest PR uses the same footer.
woodie commentedon Jun 5, 2026
There was a bit to unwind from the first implementation, but change looks solid now.
woodie commentedon Jun 5, 2026
I'm also working on simplifying this change in this PR.
Huaian666 commentedon Jun 5, 2026
one of those rare projects where the more you use it the more you appreciate it. small things keep surprising me in a good way.
woodie commentedon Jun 5, 2026
If Ginko responds to multiple flags, the current
-fdbehavior may bypasses it.• How should the
-fdbehave withreport.RunningInParallelwhich is currently ignored?• Should the logic of
DidRun()be broken out into new functions or shoulddidRunFd()be embedded intoDidRun()?I've closed PR1669 as PR1670 seems so much better to me.
onsi commentedon Jun 7, 2026
@woodie I think it's fine having this bypass other flags and essentially be a stand-alone opt-in mode. it occurs to me that it is unlikely to play well with
-randomize-alland parallel specs as in both cases there is no guarantee that tests at different container nestings won't interleave.woodie commentedon Jun 7, 2026
let me know if you want to see any other changes.
woodie commentedon Jun 22, 2026
The CHANGELOG could include something like this.
+
-fd(RSpec-style "format documentation" output) is now exclusive of-p/-procsand-randomize-all. Its hierarchical tree assumes specs run serially in declaration order, so Ginkgo now overrides those flags and prints a notice when it does, rather than producing fragmented, hard-to-read output.