Qualify objective() in its own tests, which ROI masks in a suite run

[sstoeckl]

tests/testthat/test-objective-branches.R passes when it is run on its own and
fails when the suite is run. It is not the package: by the time the file runs,
objective() is a different function.

ROI exports objective() as well. Twenty-seven of the test files call
library(ROI), and library() does not move an already-attached package back
up the search path, so a later library(PortfolioAnalytics) does not restore
it. The tests then call ROI's objective(), which takes a single argument x:

Error in objective(name = "mean", arguments = list()) :
  unused arguments (name = "mean", arguments = list())

Measured

Same file, same package build, one library() call apart:

file	library(ROI) first	PASS	FAIL	ERROR
master	no	33	0	0
master	yes	30	0	2
this branch	yes	33	0	0
this branch	no	33	0	0

The change

Three calls become PortfolioAnalytics::objective(...), with a comment saying
why, so the file tests the function it is written for regardless of what a
previous file attached.

The expected messages are pinned at the same time, because one of the three
tests passed under masking without testing anything: it expected the regexp
"name", and name also appears in R's own unused arguments (name = NULL, arguments = list()). It now expects the sentence the function actually raises.

Nothing in R/ changes.

Test suite

Run against master's package code, so the only difference is this file:

	PASS	FAIL	ERROR
master	3550	2	2
this branch	3553	2	0

No errors left. The two failures are the chart.EfficientFrontierCompare()
tests, which #60 fixes; measured there, that file goes from 67 passing
with 2 failures to 69 with none. The two changes touch different files, so
between them the four problems run-all.R reports are accounted for.

Why this one and #60 are worth taking first

R-CMD-check and test-coverage have both been red on master since 24 July,
and both die at the same step, running the tests in 'run-all.R' failed. What
run-all.R reports there is four problems: the two in this file, and two in
test-charts-roi.R that come from a missing initialisation in
chart.EfficientFrontierCompare(). #60 fixes those.

With both merged, that step passes and the pipeline is green again. I can only
see that the workflows stop at that step, so if something else is broken behind
it, it is hidden by these four rather than absent.

This matters for the five pull requests that are already open, because their
checks are red for this reason and not for theirs. They do not need rebasing:
the workflows run on pull_request and actions/checkout@v4 checks out the
merge commit, so once master carries these two fixes, re-running the jobs on
each of them is enough.

Scope

One test file, 18 insertions, 5 deletions, of which 12 insertions are the
comment. This change stands on its own.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The reviewed test changes address the suite-order failure without unresolved issues.

Pull request overview

This PR makes objective branch tests independent of ROI package masking.

Changes:

• Qualifies calls as PortfolioAnalytics::objective().
• Pins expected error messages.
• Documents the package masking issue.

File summaries

File	Description
tests/testthat/test-objective-branches.R	Ensures objective tests consistently exercise the intended function.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[braverock]

reviewed and merged. Thank you for the contribution to PortfolioAnalytics @sstoeckl !