From: Stefan Beller <sbeller@google.com>
To: gitster@pobox.com, git@vger.kernel.org, sunshine@sunshineco.com,
jacob.keller@gmail.com
Cc: norio.nomura@gmail.com, Stefan Beller <sbeller@google.com>
Subject: [PATCHv3 0/2] Fix relative path issues in recursive submodules.
Date: Thu, 31 Mar 2016 17:17:27 -0700 [thread overview]
Message-ID: <1459469849-9643-1-git-send-email-sbeller@google.com> (raw)
Thanks Junio for review!
v3:
* This is a resend of the last two patches of the series, i.e. it replaces
44859d6626d4 and efdef1e2e in sb/submodule-helper-clone-regression-fix
* use absolute_path for sm_gitdir
* removed todo
* we need to free the intermediate sm_gitdir, so rename that to sm_gitdir_rel
and free it afterwards.
* while we currently do not support an absolute `path`, we eventually might.
If `path` is absolute it would be a pointer to `argv`, which would lead to a
crash. Duplicate the path and the crash is prevented.
(* I thought we could use it as well for `path`, but we cannot; as
get_git_work_tree() != cwd)
* diff to sb/submodule-helper-clone-regression-fix:
diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c
index 89cbbda..be7bf5f 100644
--- a/builtin/submodule--helper.c
+++ b/builtin/submodule--helper.c
@@ -153,12 +153,13 @@ static int clone_submodule(const char *path, const char *gitdir, const char *url
static int module_clone(int argc, const char **argv, const char *prefix)
{
- const char *path = NULL, *name = NULL, *url = NULL;
+ const char *name = NULL, *url = NULL;
const char *reference = NULL, *depth = NULL;
int quiet = 0;
FILE *submodule_dot_git;
- char *sm_gitdir, *p;
- struct strbuf rel_path = STRBUF_INIT; /* for relative_path */
+ char *sm_gitdir_rel, *p, *path = NULL;
+ const char *sm_gitdir;
+ struct strbuf rel_path = STRBUF_INIT;
struct strbuf sb = STRBUF_INIT;
struct option module_clone_options[] = {
@@ -198,26 +199,14 @@ static int module_clone(int argc, const char **argv, const char *prefix)
die(_("submodule--helper: unspecified or empty --path"));
strbuf_addf(&sb, "%s/modules/%s", get_git_dir(), name);
- sm_gitdir = strbuf_detach(&sb, NULL);
-
-
- if (!is_absolute_path(sm_gitdir)) {
- char *cwd = xgetcwd();
- strbuf_addf(&sb, "%s/%s", cwd, sm_gitdir);
- sm_gitdir = strbuf_detach(&sb, 0);
- free(cwd);
- }
+ sm_gitdir_rel = strbuf_detach(&sb, NULL);
+ sm_gitdir = absolute_path(sm_gitdir_rel);
if (!is_absolute_path(path)) {
- /*
- * TODO: add prefix here once we allow calling from non root
- * directory?
- */
- strbuf_addf(&sb, "%s/%s",
- get_git_work_tree(),
- path);
+ strbuf_addf(&sb, "%s/%s", get_git_work_tree(), path);
path = strbuf_detach(&sb, 0);
- }
+ } else
+ path = xstrdup(path);
if (!file_exists(sm_gitdir)) {
if (safe_create_leading_directories_const(sm_gitdir) < 0)
@@ -240,6 +229,7 @@ static int module_clone(int argc, const char **argv, const char *prefix)
submodule_dot_git = fopen(sb.buf, "w");
if (!submodule_dot_git)
die_errno(_("cannot open file '%s'"), sb.buf);
+
fprintf_or_die(submodule_dot_git, "gitdir: %s\n",
relative_path(sm_gitdir, path, &rel_path));
if (fclose(submodule_dot_git))
@@ -255,8 +245,8 @@ static int module_clone(int argc, const char **argv, const char *prefix)
relative_path(path, sm_gitdir, &rel_path));
strbuf_release(&sb);
strbuf_release(&rel_path);
- free(sm_gitdir);
-
+ free(sm_gitdir_rel);
+ free(path);
free(p);
return 0;
}
v2:
* reworded commit message for test (Thanks Junio!)
* instead of "simplifying" the path handling in patch 2, we are prepared
for anything the user throws at us (i.e. instead of segfault we
die(_("submodule--helper: unspecified or empty --path"));
(Thanks Eric, Jacob for pushing back here!)
* reword commit message for patch 3 (safe_create_leading_directories_const is
not a check, Thanks Junio!)
* patch 4 (the fix): That got reworked completely. No flow controlling
conditions in the write out phase!
* patch 5 is a bonus! (replace fprintf by fprintf_or die) When implementing
that I wondered if we also have close_or_die and open_or_die, but that doesn't
seem to be the case.
Thanks,
Stefan
v1:
It took me longer than expected to fix this bug.
It comes with a test and minor refactoring and applies on top of
origin/sb/submodule-helper, such that it can be merged into 2.7, 2.8 as well
as master.
Patch 1 is a test which fails; it has a similar layout as the
real failing repository Norio Nomura pointed out, but simplified to the
bare essentials for reproduction of the bug.
Patch 2&3 are not strictly necessary for fixing the isseu, but it removes
stupid code I wrote, so the resulting code looks a little better.
Patch 4 fixes the issue by giving more information to relative_path,
such that a relative path can be found in all cases.
Thanks,
Stefan
Stefan Beller (2):
submodule--helper, module_clone: always operate on absolute paths
submodule--helper, module_clone: catch fprintf failure
builtin/submodule--helper.c | 32 ++++++++++++++------------------
t/t7400-submodule-basic.sh | 2 +-
2 files changed, 15 insertions(+), 19 deletions(-)
--
2.5.0.262.g0ef869b.dirty
next reply other threads:[~2016-04-01 0:17 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-01 0:17 Stefan Beller [this message]
2016-04-01 0:17 ` [PATCH 1/2] submodule--helper, module_clone: always operate on absolute paths Stefan Beller
2016-04-01 19:18 ` Junio C Hamano
2016-04-01 19:30 ` Junio C Hamano
2016-04-01 20:09 ` Eric Sunshine
2016-04-01 20:39 ` Junio C Hamano
2016-04-01 0:17 ` [PATCH 2/2] submodule--helper, module_clone: catch fprintf failure Stefan Beller
2016-04-01 14:41 ` [PATCHv3 0/2] Fix relative path issues in recursive submodules Ramsay Jones
2016-04-12 15:58 ` Stefan Beller
2016-04-13 19:21 ` Ramsay Jones
2016-04-13 20:34 ` Stefan Beller
2016-04-13 22:23 ` Junio C Hamano
2016-04-13 22:30 ` Stefan Beller
2016-04-13 22:45 ` Junio C Hamano
2016-04-13 21:03 ` Junio C Hamano
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1459469849-9643-1-git-send-email-sbeller@google.com \
--to=sbeller@google.com \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=jacob.keller@gmail.com \
--cc=norio.nomura@gmail.com \
--cc=sunshine@sunshineco.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).