tests/test-basic.t
10 ↗	(On Diff #26194)	You can use `(rhg !)` for this. Alternatively, we could also set it since if does not use normal `hg` invocation.
tests/test-globalopts.t
267–271	Why do we need a fallback here ?
tests/test-hgrc.t
62	What's the possible difference here ?
tests/test-ssh.t
406–423	I don't think the behavior in the alternative block is correct. It look like the think above is trying to flag some potential security issue. I recommend we don't record the behavior of `rhg` not doing it and instead keep the if with a comment to flag this for future investigation. (This is fine if rhg does not handling this yet)

tests/test-hgrc.t
62	Maybe it's the same as this issue I had a while ago: https://www.mercurial-scm.org/repo/hg/file/74e2256a56b8/tests/common-pattern.py#l126

tests/test-basic.t
10 ↗	(On Diff #26194)	I’ve seen other tests do something similar, but it’s not documented in https://www.mercurial-scm.org/wiki/WritingTests#Filtering_output and I filed to find the code implementing it in `run-tests.py` so I didn’t know how to use it. Does it mark a line that is there when that feature is enabled or disabled? Is `!` a negation? I considered always setting this but the current `run-tests.py` code doesn’t have a meaningful value when in non-rhg mode.
tests/test-globalopts.t
267–271	Whatever error causes a Python traceback (either inexistent cwd or config parse error I suppose?) goes through Rust code instead when in rhg mode, so there is no traceback. `chg` already had the same behavior so that seemed reasonable.
tests/test-hgrc.t
62	Right, the `Display` impl for `io::Error` prints the numeric errno value next to the error string: https://github.com/rust-lang/rust/blob/1.50.0/library/std/src/io/error.rs#L538-L541 so the output is `(Permission denied (os error 13))`
tests/test-ssh.t
406–423	Yes, `abort: potentially unsafe serve` indeed blocks a security issue in the `serve` command. As far as I can tell the current rhg behavior is safe since it aborts (although with a different message) before getting anywhere near the `serve`. Keeping the `if` but removing the `else` simply reduces test coverage. Having a test expect the current behavior even if it’s not the "ideal" behavior seems preferable to me as it lets us find out if a patch changes it accidentally.

tests/test-ssh.t
406–423	Should we use `(missing-correct-output !)` `(known-bad-output !)` there then ?

Diff	ID	Description	Created	Lint	Unit
Base		Base
Diff 1	26174		Mar 9 2021, 4:40 AM	★	★
Diff 2	26194		Mar 9 2021, 8:40 AM	★	★
Diff 3	26296		Mar 12 2021, 5:07 PM	★	★
Diff 4	26341		Mar 15 2021, 9:10 AM	★	★
Diff 5	26418	rHGe96a0a53de20700beacccb656617bceab78f639e	Mar 5 2021, 5:39 AM	★	★

Commit	Parents	Author	Summary	Date
79f60db5f75e	b0df140ead01	Simon Sapin		Mar 5 2021, 5:39 AM

Status	Author	Revision
Closed	SimonSapin	D10196 ci: Add a job testing with rhg installed as `hg`
Closed	SimonSapin	D10195 run-test: install rhg if --rhg is passed
Closed	SimonSapin	D10194 makefile: add a install option
Closed	SimonSapin	D10193 makefile: add a build-rhg option
Closed	SimonSapin	D10192 makefile: add a install-chg option
Closed	SimonSapin	D10191 makefile: add a build-chg option
Closed	SimonSapin	D10190 tests: Disable for rhg remaining tests that fail in that mode
Closed	SimonSapin	D10189 rhg: Remove `rhg.fallback-executable=hg` default configuration
Closed	SimonSapin	D10145 rhg: Return an error code for `rhg config Section.idontexist`
Closed	SimonSapin	D10144 rhg: Fall back to Python for --version
Closed	SimonSapin	D10143 rhg: Exit with an error code if `files` finds nothing
Closed	SimonSapin	D10142 rhg: `cat` command: print error messages for missing files
Closed	SimonSapin	D10141 rhg: Silently ignore missing files in config %include
Closed	SimonSapin	D10140 rhg: Add support for environment variables in config include paths
Closed	SimonSapin	D10139 rhg: Add support for the HGRCSKIPREPO environment variable
Closed	SimonSapin	D10138 rhg: Don’t make repository path absolute too early
Closed	SimonSapin	D10137 rhg: Fall back to Python if ui.relative-paths is configured
Closed	SimonSapin	D10136 tests: clarify some missing output in test-merge-subrepos
Closed	SimonSapin	D10135 tests: Adapt expected output for minor differences with rhg
Closed	SimonSapin	D10134 rhg: Add support for --cwd
Closed	SimonSapin	D10133 rhg: Align with Python on some more error messages
Closed	SimonSapin	D10188 rhg: Add an allow-list of ignored extensions
Closed	SimonSapin	D10132 rhg: Ignore trailing newlines in .hg/sharedpath
Closed	SimonSapin	D10112 rhg: Fall back to Python if unsupported extensions are enabled
Closed	SimonSapin	D10111 rhg: Sort config files when adding a directory
Closed	SimonSapin	D10110 rhg: Align config file parse error formatting with Python
Closed	SimonSapin	D10102 rhg: Fall back to Python for bundle repositories
Closed	SimonSapin	D10101 rhg: Fall back to Python on --repository with an URL
Closed	SimonSapin	D10100 rhg: Print non-absolutized path in "repository {} not found" errors
Closed	SimonSapin	D10099 rhg: Align "malformed --config" error message with Python
Closed	SimonSapin	D10098 rhg: Fall back to Python on unsupported `rhg config <section>`
Closed	SimonSapin	D10097 rhg: Fall back to Python for unsupported revset syntax
Closed	SimonSapin	D10096 tests: Add `rhg` and `no-rhg` for #require and #if in .t files
Closed	SimonSapin	D10187 tests: Enable rhg fallback to Python by default in tests
Closed	SimonSapin	D10186 rhg: Add environment variables for fallback configuration

Diff 26341

tests/test-globalopts.t

	extra: branch=default			extra: branch=default
	description:			description:
	a			a



	Testing --traceback:			Testing --traceback:

	#if no-chg			#if no-chg no-rhg
	$ hg --cwd c --config x --traceback id 2>&1 \| grep -i 'traceback'			$ hg --cwd c --config x --traceback id 2>&1 \| grep -i 'traceback'
	Traceback (most recent call last):			Traceback (most recent call last):
	Traceback (most recent call last): (py3 !)			Traceback (most recent call last): (py3 !)
	#else			#else
				marmouteUnsubmitted Not Done Why do we need a fallback here ? marmoute: Why do we need a fallback here ?
				SimonSapinAuthorUnsubmitted Done Whatever error causes a Python traceback (either inexistent cwd or config parse error I suppose?) goes through Rust code instead when in rhg mode, so there is no traceback. `chg` already had the same behavior so that seemed reasonable. SimonSapin: Whatever error causes a Python traceback (either inexistent cwd or config parse error I suppose?
	Traceback for '--config' errors not supported with chg.			Traceback for '--config' errors not supported with chg.
	$ hg --cwd c --config x --traceback id 2>&1 \| grep -i 'traceback'			$ hg --cwd c --config x --traceback id 2>&1 \| grep -i 'traceback'
	[1]			[1]
	#endif			#endif

	Testing --time:			Testing --time:

	$ hg --cwd a --time id			$ hg --cwd a --time id

tests/test-hgrc.t

	$ echo '[section]' > $TESTTMP/included			$ echo '[section]' > $TESTTMP/included
	$ echo 'option = value' >> $TESTTMP/included			$ echo 'option = value' >> $TESTTMP/included
	$ echo '%include $TESTTMP/included' >> $HGRC			$ echo '%include $TESTTMP/included' >> $HGRC
	$ hg showconfig section			$ hg showconfig section
	section.option=value			section.option=value
	#if unix-permissions no-root			#if unix-permissions no-root
	$ chmod u-r $TESTTMP/included			$ chmod u-r $TESTTMP/included
	$ hg showconfig section			$ hg showconfig section
	config error at $TESTTMP/hgrc:2: cannot include $TESTTMP/included (Permission denied)			config error at $TESTTMP/hgrc:2: cannot include $TESTTMP/included (Permission denied*) (glob)
				marmouteUnsubmitted Not Done What's the possible difference here ? marmoute: What's the possible difference here ?
				AlphareUnsubmitted Not Done Maybe it's the same as this issue I had a while ago: https://www.mercurial-scm.org/repo/hg/file/74e2256a56b8/tests/common-pattern.py#l126 Alphare: Maybe it's the same as this issue I had a while ago: https://www.mercurial-scm.
				SimonSapinAuthorUnsubmitted Done Right, the `Display` impl for `io::Error` prints the numeric errno value next to the error string: https://github.com/rust-lang/rust/blob/1.50.0/library/std/src/io/error.rs#L538-L541 so the output is `(Permission denied (os error 13))` SimonSapin: Right, the `Display` impl for `io::Error` prints the numeric errno value next to the error…
	[255]			[255]
	#endif			#endif

	issue1829: wrong indentation			issue1829: wrong indentation

	$ echo '[foo]' > $HGRC			$ echo '[foo]' > $HGRC
	$ echo ' x = y' >> $HGRC			$ echo ' x = y' >> $HGRC
	$ hg version			$ hg version

tests/test-ssh.t


	Test (non-)escaping of remote paths with spaces when cloning (issue3145):			Test (non-)escaping of remote paths with spaces when cloning (issue3145):

	$ hg clone --ssh "\"$PYTHON\" \"$TESTDIR/dummyssh\"" "ssh://user@dummy/a repo"			$ hg clone --ssh "\"$PYTHON\" \"$TESTDIR/dummyssh\"" "ssh://user@dummy/a repo"
	destination directory: a repo			destination directory: a repo
	abort: destination 'a repo' is not empty			abort: destination 'a repo' is not empty
	[10]			[10]

				#if no-rhg
	Make sure hg is really paranoid in serve --stdio mode. It used to be			Make sure hg is really paranoid in serve --stdio mode. It used to be
	possible to get a debugger REPL by specifying a repo named --debugger.			possible to get a debugger REPL by specifying a repo named --debugger.
	$ hg -R --debugger serve --stdio			$ hg -R --debugger serve --stdio
	abort: potentially unsafe serve --stdio invocation: ['-R', '--debugger', 'serve', '--stdio']			abort: potentially unsafe serve --stdio invocation: ['-R', '--debugger', 'serve', '--stdio']
	[255]			[255]
	$ hg -R --config=ui.debugger=yes serve --stdio			$ hg -R --config=ui.debugger=yes serve --stdio
	abort: potentially unsafe serve --stdio invocation: ['-R', '--config=ui.debugger=yes', 'serve', '--stdio']			abort: potentially unsafe serve --stdio invocation: ['-R', '--config=ui.debugger=yes', 'serve', '--stdio']
	[255]			[255]
	Abbreviations of 'serve' also don't work, to avoid shenanigans.			Abbreviations of 'serve' also don't work, to avoid shenanigans.
	$ hg -R narf serv --stdio			$ hg -R narf serv --stdio
	abort: potentially unsafe serve --stdio invocation: ['-R', 'narf', 'serv', '--stdio']			abort: potentially unsafe serve --stdio invocation: ['-R', 'narf', 'serv', '--stdio']
	[255]			[255]
				#else
				rhg aborts early on -R without a repository at that path
				$ hg -R --debugger serve --stdio
				abort: potentially unsafe serve --stdio invocation: ['-R', '--debugger', 'serve', '--stdio'] (missing-correct-output !)
				abort: repository --debugger not found (known-bad-output !)
				[255]
				$ hg -R --config=ui.debugger=yes serve --stdio
				abort: potentially unsafe serve --stdio invocation: ['-R', '--config=ui.debugger=yes', 'serve', '--stdio'] (missing-correct-output !)
				abort: repository --config=ui.debugger=yes not found (known-bad-output !)
				[255]
				$ hg -R narf serv --stdio
				abort: potentially unsafe serve --stdio invocation: ['-R', 'narf', 'serv', '--stdio'] (missing-correct-output !)
				abort: repository narf not found (known-bad-output !)
				[255]
				If the repo does exist, rhg finds an unsupported command and falls back to Python
				which still does the right thing
				$ hg init narf
				$ hg -R narf serv --stdio
				marmouteUnsubmitted Not Done I don't think the behavior in the alternative block is correct. It look like the think above is trying to flag some potential security issue. I recommend we don't record the behavior of `rhg` not doing it and instead keep the if with a comment to flag this for future investigation. (This is fine if rhg does not handling this yet) marmoute: I don't think the behavior in the alternative block is correct. It look like the think above is…
				SimonSapinAuthorUnsubmitted Done Yes, `abort: potentially unsafe serve` indeed blocks a security issue in the `serve` command. As far as I can tell the current rhg behavior is safe since it aborts (although with a different message) before getting anywhere near the `serve`. Keeping the `if` but removing the `else` simply reduces test coverage. Having a test expect the current behavior even if it’s not the "ideal" behavior seems preferable to me as it lets us find out if a patch changes it accidentally. SimonSapin: Yes, `abort: potentially unsafe serve` indeed blocks a security issue in the `serve` command.
				marmouteUnsubmitted Not Done Should we use `(missing-correct-output !)` `(known-bad-output !)` there then ? marmoute: Should we use `(missing-correct-output !)` `(known-bad-output !)` there then ?
				abort: potentially unsafe serve --stdio invocation: ['-R', 'narf', 'serv', '--stdio']
				[255]
				#endif

	Test hg-ssh using a helper script that will restore PYTHONPATH (which might			Test hg-ssh using a helper script that will restore PYTHONPATH (which might
	have been cleared by a hg.exe wrapper) and invoke hg-ssh with the right			have been cleared by a hg.exe wrapper) and invoke hg-ssh with the right
	parameters:			parameters:

	$ cat > ssh.sh << EOF			$ cat > ssh.sh << EOF
	> userhost="\$1"			> userhost="\$1"
	> SSH_ORIGINAL_COMMAND="\$2"			> SSH_ORIGINAL_COMMAND="\$2"

			Path	Packages
M			tests/test-globalopts.t (2 lines)
M			tests/test-hgrc.t (2 lines)
M			tests/test-ssh.t (22 lines)

This is an archive of the discontinued Mercurial Phabricator instance.

tests: Adapt expected output for minor differences with rhg
ClosedPublic

Details

Diff Detail

Event Timeline

Revision Contents
Changeset List

Diff 26341

tests/test-globalopts.t

tests/test-hgrc.t

tests/test-ssh.t

This is an archive of the discontinued Mercurial Phabricator instance.

tests: Adapt expected output for minor differences with rhgClosedPublic

Details

Diff Detail

Event Timeline

Revision ContentsChangeset List

Diff 26341

tests/test-globalopts.t

tests/test-hgrc.t

tests/test-ssh.t

tests: Adapt expected output for minor differences with rhg
ClosedPublic

Revision Contents
Changeset List