From: Andrzej Hunt Date: Wed, 14 Jul 2021 18:57:37 GMT Subject: Re: [PATCH v2 2/4] SANITIZE tests: fix memory leaks in t13*config*, add to whitelist Message-ID: <871ea493-e108-e748-0234-f929690ad2fd@ahunt.org> In-Reply-To: On 14/07/2021 19:23, Ævar Arnfjörð Bjarmason wrote: > Fix a couple of trivial memory leaks introduced in 3efd0bedc6 (config: > add conditional include, 2017-03-01) and my own 867ad08a26 (hooks: > allow customizing where the hook directory is, 2016-05-04). > > In the latter case the "fix" is UNLEAK() on the global variable. This > allows us to run all t13*config* tests under SANITIZE=leak. > > With this change we can now run almost the whole set of config.c > tests (t13*config) under SANITIZE=leak, so let's do so, with a few > exceptions: > > * The test added in ce81b1da23 (config: add new way to pass config > via `--config-env`, 2021-01-12), it fails in GitHub CI, but passes > for me locally. Let's just skip it for now. > > * Ditto the split_cmdline and "aliases of builtins" tests, the former > required splitting up an existing test, there an issue with the test > that would have also been revealed by skipping it. > > Signed-off-by: Ævar Arnfjörð Bjarmason > --- > config.c | 17 ++++++++++++----- > t/t1300-config.sh | 16 ++++++++++------ > t/test-lib.sh | 1 + > 3 files changed, 23 insertions(+), 11 deletions(-) > > diff --git a/config.c b/config.c > index f9c400ad30..38e132c0e2 100644 > --- a/config.c > +++ b/config.c > @@ -138,8 +138,10 @@ static int handle_path_include(const char *path, struct config_include_data *inc > return config_error_nonbool("include.path"); > > expanded = expand_user_path(path, 0); > - if (!expanded) > - return error(_("could not expand include path '%s'"), path); > + if (!expanded) { > + ret = error(_("could not expand include path '%s'"), path); > + goto cleanup; > + } > path = expanded; > > /* > @@ -149,8 +151,10 @@ static int handle_path_include(const char *path, struct config_include_data *inc > if (!is_absolute_path(path)) { > char *slash; > > - if (!cf || !cf->path) > - return error(_("relative config includes must come from files")); > + if (!cf || !cf->path) { > + ret = error(_("relative config includes must come from files")); > + goto cleanup; > + } > > slash = find_last_dir_sep(cf->path); > if (slash) > @@ -168,6 +172,7 @@ static int handle_path_include(const char *path, struct config_include_data *inc > ret = git_config_from_file(git_config_include, path, inc); > inc->depth--; > } > +cleanup: > strbuf_release(&buf); > free(expanded); > return ret; > @@ -1331,8 +1336,10 @@ static int git_default_core_config(const char *var, const char *value, void *cb) > if (!strcmp(var, "core.attributesfile")) > return git_config_pathname(&git_attributes_file, var, value); > > - if (!strcmp(var, "core.hookspath")) > + if (!strcmp(var, "core.hookspath")) { > + UNLEAK(git_hooks_path); > return git_config_pathname(&git_hooks_path, var, value); > + } Why is the UNLEAK necessary here? We generally want to limit use of UNLEAK to cmd_* functions or direct helpers. git_default_core_config() seems generic enough that it could be called from anywhere, and using UNLEAK here means we're potentially masking a real leak? IIUC the leak here happens because: - git_hooks_path is a global variable - hence it's unlikely we'd ever bother cleaning it up. - git_default_core_config() gets called a first time with core.hookspath, and we end up allocating new memory into git_hooks_path. - git_default_core_config() gets called again with core.hookspath, and we overwrite git_hooks_path with a new string which leaks the string that git_hooks_path used to point to. So I think the real fix is to free(git_hooks_path) instead of an UNLEAK? (Looking at the surrounding code, it looks like the same pattern of leak might be repeated for other similar globals - is it worth auditing those while we're here?) > > if (!strcmp(var, "core.bare")) { > is_bare_repository_cfg = git_config_bool(var, value); > diff --git a/t/t1300-config.sh b/t/t1300-config.sh > index 9ff46f3b04..93ad0f4887 100755 > --- a/t/t1300-config.sh > +++ b/t/t1300-config.sh > @@ -1050,12 +1050,16 @@ test_expect_success SYMLINKS 'symlink to nonexistent configuration' ' > test_must_fail git config --file=linktolinktonada --list > ' > > -test_expect_success 'check split_cmdline return' " > - git config alias.split-cmdline-fix 'echo \"' && > - test_must_fail git split-cmdline-fix && > +test_expect_success 'setup check split_cmdline return' " > echo foo > foo && > git add foo && > - git commit -m 'initial commit' && > + git commit -m 'initial commit' > +" > + > +test_expect_success !SANITIZE_LEAK 'check split_cmdline return' " > + git config alias.split-cmdline-fix 'echo \"' && > + test_must_fail git split-cmdline-fix && > + > git config branch.main.mergeoptions 'echo \"' && > test_must_fail git merge main > " > @@ -1101,7 +1105,7 @@ test_expect_success 'key sanity-checking' ' > git config foo."ba =z".bar false > ' > > -test_expect_success 'git -c works with aliases of builtins' ' > +test_expect_success !SANITIZE_LEAK 'git -c works with aliases of builtins' ' > git config alias.checkconfig "-c foo.check=bar config foo.check" && > echo bar >expect && > git checkconfig >actual && > @@ -1397,7 +1401,7 @@ test_expect_success 'git --config-env with missing value' ' > grep "invalid config format: config" error > ' > > -test_expect_success 'git --config-env fails with invalid parameters' ' > +test_expect_success !SANITIZE_LEAK 'git --config-env fails with invalid parameters' ' > test_must_fail git --config-env=foo.flag config --bool foo.flag 2>error && > test_i18ngrep "invalid config format: foo.flag" error && > test_must_fail git --config-env=foo.flag= config --bool foo.flag 2>error && > diff --git a/t/test-lib.sh b/t/test-lib.sh > index 9201510e16..98e20950c3 100644 > --- a/t/test-lib.sh > +++ b/t/test-lib.sh > @@ -1370,6 +1370,7 @@ maybe_skip_all_sanitize_leak () { > add_sanitize_leak_true 't000*' > add_sanitize_leak_true 't001*' > add_sanitize_leak_true 't006*' > + add_sanitize_leak_true 't13*config*' > > # Blacklist patterns (overrides whitelist) > add_sanitize_leak_false 't000[469]*' >