pypa/pip · error · ValueError

Badly formatted data

Error message

Badly formatted data: {data!r}

What it means

Raised by Subversion._get_svn_url_rev() when parsing the legacy SVN 'entries' file in XML format and the repository URL regex (_svn_xml_url_re) finds no match. This indicates the SVN metadata file exists and starts with '<?xml' but does not contain a recognizable url attribute.

Solutions

1. Upgrade the SVN client: 'svn --version' — modern SVN uses a format that bypasses this code path
2. Re-checkout the working copy with a current SVN client: 'svn checkout <url>'
3. Run 'svn info <location>' manually to verify the working copy is valid
4. Avoid SVN-based pip installs if possible; migrate to git or use a direct download

Example fix

// before
pip install svn+http://example.com/svn/MyApp#egg=MyApp
# old SVN working copy with malformed entries
// after
svn --version  # upgrade if < 1.7
svn checkout http://example.com/svn/MyApp fresh_checkout
pip install ./fresh_checkout

import re

_XML_URL_RE = re.compile('url="([^"]+)"')

def svn_xml_has_url(data: str) -> bool:
    return bool(_XML_URL_RE.search(data)) if data.startswith('<?xml') else True

try:
    url, rev = Subversion._get_svn_url_rev(location)
except ValueError as e:
    if 'Badly formatted data' in str(e):
        # re-checkout with a modern SVN client
        raise EnvironmentError(f'SVN metadata corrupt or unsupported version: {e}')
    raise

When it happens

Trigger: Calling _get_svn_url_rev(location) where the .svn/entries file starts with '<?xml' but _svn_xml_url_re.search(data) returns None (no url="..." attribute found). Reached during SVN-based VCS installation or when pip inspects an SVN working copy.

Common situations: A very old or unusual SVN client version that wrote the entries file in a different XML schema. A corrupted .svn/entries file. An SVN working copy checked out by a third-party tool. This is increasingly rare as modern SVN (>= 1.7) uses a different metadata format that falls through to 'svn info'.

Related errors

• Error parsing entry points for
• Error parsing entry points for
• Ill-formed name/version string
• Invalid requirement
• Invalid section heading

AI-assisted analysis of pypa/pip@f399c37189 (2026-08-08). Data as JSON: /api/errors/582f2b8d02656538. Report an issue: GitHub.

Appendix: source

Thrown at src/pip/_internal/vcs/subversion.py:151

        from pip._internal.exceptions import InstallationError

        entries_path = os.path.join(location, cls.dirname, "entries")
        if os.path.exists(entries_path):
            with open(entries_path) as f:
                data = f.read()
        else:  # subversion >= 1.7 does not have the 'entries' file
            data = ""

        url = None
        if data.startswith(("8", "9", "10")):
            entries = list(map(str.splitlines, data.split("\n\x0c\n")))
            del entries[0][0]  # get rid of the '8'
            url = entries[0][3]
            revs = [int(d[9]) for d in entries if len(d) > 9 and d[9]] + [0]
        elif data.startswith("<?xml"):
            match = _svn_xml_url_re.search(data)
            if not match:
                raise ValueError(f"Badly formatted data: {data!r}")
            url = match.group(1)  # get repository URL
            revs = [int(m.group(1)) for m in _svn_rev_re.finditer(data)] + [0]
        else:
            try:
                # subversion >= 1.7
                # Note that using get_remote_call_options is not necessary here
                # because `svn info` is being run against a local directory.
                # We don't need to worry about making sure interactive mode
                # is being used to prompt for passwords, because passwords
                # are only potentially needed for remote server requests.
                xml = cls.run_command(
                    ["info", "--xml", location],
                    show_stdout=False,
                    stdout_only=True,
                )
                match = _svn_info_xml_url_re.search(xml)
                assert match is not None
                url = match.group(1)

View on GitHub (pinned to f399c37189)