A couple errors dealing with uninitialized submodules

2 messages, 2 authors, 2016-10-20 · open the first message on its own page

A couple errors dealing with uninitialized submodules

From: Aaron Schrab <hidden>
Date: 2016-10-20 23:36:20

I was working with a fresh clone of a repository where I'd forgotten 
that one of the directories was setup as a submodule, so I hadn't 
initialized it.

I tried to add a symlink to a location outside the repository and this 
failed with an assertion (exact text in comment below). When looking 
into that I realized that the directory was meant to be a submodule and 
decided to try to change that.  So I tried to remove the submodule, and 
got an error (misleadingly) saying that couldn't be done because it uses 
a .git directory.

I first noticed this with git 2.9.3 from Debian unstable, but I also see 
it building from v2.10.1-502-g6598894 fetched from master recently.

The following script replicates both of these issues. These could both 
be classified as "don't do that", although at lease the assertion is 
quite ugly.
--- >8 ----
#!/bin/sh -e

directory=$(mktemp -d)
echo "Using directory '$directory'"
cd $directory
git init --quiet orig
(
  cd orig
  # Using a random, small repository for the submodule.
  git submodule --quiet add https://github.com/diepm/vim-rest-console.git sub
  git commit -m init >/dev/null
)
git clone --quiet orig dup
cd dup

(
  cd sub
  ln -s /tmp/dont_care
  # Next command aborts with
  # git: pathspec.c:317: prefix_pathspec: Assertion `item->nowildcard_len <= item->len && item->prefix <= item->len' failed.`
  git add . || : expected to fail
)

rm -f .git/index.lock
# Next command fails with error wrongly saying that the submodule uses a .git
# directory.  I believe that the real problem is that the uninitialized
# submodule has content.
git rm sub

Re: A couple errors dealing with uninitialized submodules

From: Stefan Beller <hidden>
Date: 2016-10-20 23:51:13

On Thu, Oct 20, 2016 at 4:29 PM, Aaron Schrab [off-list ref] wrote:
quoted hunk
I was working with a fresh clone of a repository where I'd forgotten that
one of the directories was setup as a submodule, so I hadn't initialized it.

I tried to add a symlink to a location outside the repository and this
failed with an assertion (exact text in comment below). When looking into
that I realized that the directory was meant to be a submodule and decided
to try to change that.  So I tried to remove the submodule, and got an error
(misleadingly) saying that couldn't be done because it uses a .git
directory.

I first noticed this with git 2.9.3 from Debian unstable, but I also see it
building from v2.10.1-502-g6598894 fetched from master recently.

The following script replicates both of these issues. These could both be
classified as "don't do that", although at lease the assertion is quite
ugly.
--- >8 ----
#!/bin/sh -e

directory=$(mktemp -d)
echo "Using directory '$directory'"
cd $directory
git init --quiet orig
(
 cd orig
 # Using a random, small repository for the submodule.
 git submodule --quiet add https://github.com/diepm/vim-rest-console.git sub
 git commit -m init >/dev/null
)
git clone --quiet orig dup
cd dup

(
 cd sub
 ln -s /tmp/dont_care
 # Next command aborts with
 # git: pathspec.c:317: prefix_pathspec: Assertion `item->nowildcard_len <=
item->len && item->prefix <= item->len' failed.`
For this bug see
https://public-inbox.org/git/CAFOYHZDw-P0ST8WKoSVxBpbFCiACZpgiDPMfw5MRtFTMosO0rg@mail.gmail.com/
Specifically try this patch
https://public-inbox.org/git/CAGZ79kZaZCwZ-cesB_voq0s0Qt+ipcgb6TkdzLE+EWSF_qRj7A@mail.gmail.com/
 git add . || : expected to fail
)

rm -f .git/index.lock
# Next command fails with error wrongly saying that the submodule uses a
.git
# directory.  I believe that the real problem is that the uninitialized
# submodule has content.
git rm sub
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help