+sub test_getsetup {
+ # anything worth testing?
+}
+
+sub test_cvs_info {
+ # inspect "Repository revision" (used in code)
+ # inspect "Sticky Options" (used in tests to verify existence of "-kb")
+}
+
+sub test_cvs_run_cvs {
+ # extract the stdout-redirect thing
+ # - prove that it silences stdout
+ # - prove that stderr comes through just fine
+ # prove that when cvs exits nonzero (fail), function exits false
+ # prove that when cvs exits zero (success), function exits true
+ # always pass -f, just in case
+ # steal from git.pm: safe_git(), run_or_{die,cry,non}
+ # - open() instead of system()
+ # always call cvs_run_cvs(), don't ever run 'cvs' directly
+}
+
+sub test_cvs_run_cvsps {
+ # parameterize command like run_cvs()
+ # expose config vars for e.g. "--cvs-direct -z 30"
+ # always pass -x (unless proven otherwise)
+ # always pass -b HEAD (configurable like gitmaster_branch?)
+}
+
+sub test_cvs_parse_cvsps {
+ # extract method from rcs_recentchanges
+ # document expected changeset format
+ # document expected changeset delimiter
+ # try: cvsps -q -x -p && ls | sort -rn | head -100
+ # - benchmark against current impl (that uses File::ReadBackwards)
+}
+
+sub test_cvs_parse_log_accum {
+ # add new, preferred method for rcs_recentchanges to use
+ # teach log_accum to record commits (into transient?)
+ # script cvsps to bootstrap (or replace?) commit history
+ # teach ikiwiki-makerepo to set up log_accum and commit_prep
+ # why are NetBSD commit mails unreliable?
+ # - is it working for CVS commits and failing for web commits?
+}
+
+sub test_cvs_is_controlling {
+ # with no args:
+ # - if srcdir is in CVS, return true
+ # - else, return false
+ # with a dir arg:
+ # - if dir is in CVS, return true
+ # - else, return false
+ # with a file arg:
+ # - is there anything that wants the answer? if so, answer
+ # - else, die
+}
+
+sub test_rcs_update {
+ # can it assume we're under CVS control? or must it check?
+ # anything else worth testing?
+}
+
+sub test_rcs_prepedit {
+ # can it assume we're under CVS control? or must it check?
+ # for existing file, returns latest revision in repo
+ # - what's this used for? should it return latest revision in checkout?
+ # for new file, returns empty string
+}
+
+sub test_rcs_commit {
+ # can it assume we're under CVS control? or must it check?
+ # if someone else changed the page since rcs_prepedit was called:
+ # - try to merge into our working copy
+ # - if merge succeeds, proceed to commit
+ # - else, return page content with the conflict markers in it
+ # commit:
+ # - if success, return undef
+ # - else, revert + return content with the conflict markers in it
+ # git.pm receives "session" param -- useful here?
+ # web commits start with "web commit {by,from} "
+ # seeing File::chdir errors on commit?
+}
+
+sub test_rcs_commit_staged {
+ # if commit succeeds, return undef
+ # else, warn and return error message (really? or just non-undef?)
+}
+
+sub test_rcs_add {
+ my $dir1 = "test3";
+ my $dir2 = "test4/test5";
+ ok(
+ mkdir($config{srcdir} . "/$dir1"),
+ qq{can make $dir1},
+ );
+ IkiWiki::rcs_add($dir1);
+ IkiWiki::rcs_commit(
+ file => $dir1,
+ message => "shouldn't happen",
+ token => "oom",
+ );
+
+ # can it assume we're under CVS control? or must it check?
+ # add a top-level text file
+ # - rcs_commit it
+ # - inspect recentchanges: new change, no -kb
+ # add a top-level dir
+ # - test mustn't hang (does it hang if we comment out genwrapper?)
+ # - inspect recentchanges: no new change
+ # - rcs_commit it
+ # - reinspect recentchanges: still no new change
+ # add a text file in that dir
+ # - rcs_commit_staged
+ # - inspect recentchanges: new change, no -kb
+ # add a top-level dir + add a binary file in it
+ # - rcs_commit_staged
+ # - inspect recentchanges: new change, yes -kb
+ # add a top-level dir + subdir + add one text and one binary file in it
+ # - rcs_commit_staged
+ # - inspect recentchanges: one new change, two files, one -kb, one not
+
+ # extract method: filetype-guessing
+ # add a binary file, remove it, add a text file by same name, no -kb?
+ # add a text file, remove it, add a binary file by same name, -kb?
+}
+
+sub test_rcs_remove {
+ # can it assume we're under CVS control? or must it check?
+ # remove a top-level file
+ # - rcs_commit
+ # - inspect recentchanges: one new change, file removed
+ # remove two files (in different dirs)
+ # - rcs_commit_staged
+ # - inspect recentchanges: one new change, both files removed
+}
+
+sub test_rcs_rename {
+ # can it assume we're under CVS control? or must it check?
+ # rename a file in the same dir
+ # - rcs_commit_staged
+ # - inspect recentchanges: one new change, one file removed, one added
+ # rename a file into a different dir
+ # - rcs_commit_staged
+ # - inspect recentchanges: one new change, one file removed, one added
+ # rename a file into a not-yet-existing dir
+ # - rcs_commit_staged
+ # - inspect recentchanges: one new change, one file removed, one added
+ # is it safe to use "mv"? what if $dest is somehow outside the wiki?
+}
+
+sub test_rcs_recentchanges {
+ # can it assume we're under CVS control? or must it check?
+ # don't worry whether we're called with a number (we always are)
+ # other rcs tests already inspect much of the returned structure
+ # CVS commits say "cvs" and get the right committer
+ # web commits say "web" and get the right committer
+ # - and don't start with "web commit {by,from} "
+ # "nickname" -- can we ever meaningfully set this?
+
+ # prefer log_accum, then cvsps, else die
+ # run the high-level recentchanges tests 2x (once for each method)
+ # - including in other test subs that check recentchanges?
+}
+
+sub test_rcs_diff {
+ # can it assume we're under CVS control? or must it check?
+ # in list context, return all lines (with \n), up to $maxlines if set
+ # in scalar context, return the whole diff, up to $maxlines if set
+}
+
+sub test_rcs_getctime {
+ # can it assume we're under CVS control? or must it check?
+ # given a file, find its creation time, else return 0
+ # first implement in the obvious way
+ # then cache
+}
+
+sub test_rcs_getmtime {
+ # can it assume we're under CVS control? or must it check?
+ # given a file, find its modification time, else return 0
+ # first implement in the obvious way
+ # then cache
+}
+
+sub test_rcs_receive {
+ pass(q{rcs_receive doesn't make sense for CVS});
+}
+
+sub test_rcs_preprevert {
+ # can it assume we're under CVS control? or must it check?
+ # given a patchset number, return structure describing what'd happen:
+ # - see doc/plugins/write.mdwn:rcs_receive()
+ # don't forget about attachments
+}
+
+sub test_rcs_revert {
+ # can it assume we're under CVS control? or must it check?
+ # given a patchset number, stage the revert for rcs_commit_staged()
+ # if commit succeeds, return undef
+ # else, warn and return error message (really? or just non-undef?)
+}
+
+sub main {
+ my $default_test_methods = '^test_*';
+ my $test_methods = defined $ENV{TEST_METHOD}
+ ? $ENV{TEST_METHOD}
+ : $default_test_methods;
+
+ _startup($test_methods eq $default_test_methods);
+ $_->() foreach get_test_subs($test_methods);
+ _shutdown($test_methods eq $default_test_methods);
+}
+
+main();
+
+################################################################################
+# don't want a dependency on Test::Class; do want a couple of its features
+
+sub get_test_subs {
+ my $re = shift;
+ no strict 'refs';
+ return map { \&{*$_} } grep { /$re/ } list_module('main');
+}
+
+# http://stackoverflow.com/questions/607282/whats-the-best-way-to-discover-all-subroutines-a-perl-module-has
+
+use B qw/svref_2object/;
+
+sub in_package {
+ my ($coderef, $package) = @_;
+ my $cv = svref_2object($coderef);
+ return if not $cv->isa('B::CV') or $cv->GV->isa('B::SPECIAL');
+ return $cv->GV->STASH->NAME eq $package;
+}
+
+sub list_module {
+ my $module = shift;
+ no strict 'refs';
+ return grep {
+ defined &{"$module\::$_"} and in_package(\&{*$_}, $module)
+ } keys %{"$module\::"};
+}