1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
//! A crate for using libcurl as a backend for HTTP git requests with git2-rs.
//!
//! This crate provides one public function, `register`, which will register
//! a custom HTTP transport with libcurl for any HTTP requests made by libgit2.
//! At this time the `register` function is unsafe for the same reasons that
//! `git2::transport::register` is also unsafe.
//!
//! It is not recommended to use this crate wherever possible. The current
//! libcurl backend used, `curl-rust`, only supports executing a request in one
//! method call implying no streaming support. This consequently means that
//! when a repository is cloned the entire contents of the repo are downloaded
//! into memory, and *then* written off to disk by libgit2 afterwards. It
//! should be possible to alleviate this problem in the future.
//!
//! > **NOTE**: At this time this crate likely does not support a `git push`
//! >           operation, only clones.

#![doc(html_root_url = "https://docs.rs/git2-curl/0.20")]
#![deny(missing_docs)]
#![warn(rust_2018_idioms)]
#![cfg_attr(test, deny(warnings))]

use std::error;
use std::io::prelude::*;
use std::io::{self, Cursor};
use std::str;
use std::sync::{Arc, Mutex, Once};

use curl::easy::{Easy, List};
use git2::transport::SmartSubtransportStream;
use git2::transport::{Service, SmartSubtransport, Transport};
use git2::Error;
use log::{debug, info};
use url::Url;

struct CurlTransport {
    handle: Arc<Mutex<Easy>>,
    /// The URL of the remote server, e.g. `https://github.com/user/repo`
    ///
    /// This is an empty string until the first action is performed.
    /// If there is an HTTP redirect, this will be updated with the new URL.
    base_url: Arc<Mutex<String>>,
}

struct CurlSubtransport {
    handle: Arc<Mutex<Easy>>,
    service: &'static str,
    url_path: &'static str,
    base_url: Arc<Mutex<String>>,
    method: &'static str,
    reader: Option<Cursor<Vec<u8>>>,
    sent_request: bool,
}

/// Register the libcurl backend for HTTP requests made by libgit2.
///
/// This function takes one parameter, a `handle`, which is used to perform all
/// future HTTP requests. The handle can be previously configured with
/// information such as proxies, SSL information, etc.
///
/// This function is unsafe largely for the same reasons as
/// `git2::transport::register`:
///
/// * The function needs to be synchronized against all other creations of
///   transport (any API calls to libgit2).
/// * The function will leak `handle` as once registered it is not currently
///   possible to unregister the backend.
///
/// This function may be called concurrently, but only the first `handle` will
/// be used. All others will be discarded.
pub unsafe fn register(handle: Easy) {
    static INIT: Once = Once::new();

    let handle = Arc::new(Mutex::new(handle));
    let handle2 = handle.clone();
    INIT.call_once(move || {
        git2::transport::register("http", move |remote| factory(remote, handle.clone())).unwrap();
        git2::transport::register("https", move |remote| factory(remote, handle2.clone())).unwrap();
    });
}

fn factory(remote: &git2::Remote<'_>, handle: Arc<Mutex<Easy>>) -> Result<Transport, Error> {
    Transport::smart(
        remote,
        true,
        CurlTransport {
            handle: handle,
            base_url: Arc::new(Mutex::new(String::new())),
        },
    )
}

impl SmartSubtransport for CurlTransport {
    fn action(
        &self,
        url: &str,
        action: Service,
    ) -> Result<Box<dyn SmartSubtransportStream>, Error> {
        let mut base_url = self.base_url.lock().unwrap();
        if base_url.len() == 0 {
            *base_url = url.to_string();
        }
        let (service, path, method) = match action {
            Service::UploadPackLs => ("upload-pack", "/info/refs?service=git-upload-pack", "GET"),
            Service::UploadPack => ("upload-pack", "/git-upload-pack", "POST"),
            Service::ReceivePackLs => {
                ("receive-pack", "/info/refs?service=git-receive-pack", "GET")
            }
            Service::ReceivePack => ("receive-pack", "/git-receive-pack", "POST"),
        };
        info!("action {} {}", service, path);
        Ok(Box::new(CurlSubtransport {
            handle: self.handle.clone(),
            service: service,
            url_path: path,
            base_url: self.base_url.clone(),
            method: method,
            reader: None,
            sent_request: false,
        }))
    }

    fn close(&self) -> Result<(), Error> {
        Ok(()) // ...
    }
}

impl CurlSubtransport {
    fn err<E: Into<Box<dyn error::Error + Send + Sync>>>(&self, err: E) -> io::Error {
        io::Error::new(io::ErrorKind::Other, err)
    }

    fn execute(&mut self, data: &[u8]) -> io::Result<()> {
        if self.sent_request {
            return Err(self.err("already sent HTTP request"));
        }
        let agent = format!("git/1.0 (git2-curl {})", env!("CARGO_PKG_VERSION"));

        // Parse our input URL to figure out the host
        let url = format!("{}{}", self.base_url.lock().unwrap(), self.url_path);
        let parsed = Url::parse(&url).map_err(|_| self.err("invalid url, failed to parse"))?;
        let host = match parsed.host_str() {
            Some(host) => host,
            None => return Err(self.err("invalid url, did not have a host")),
        };

        // Prep the request
        debug!("request to {}", url);
        let mut h = self.handle.lock().unwrap();
        h.url(&url)?;
        h.useragent(&agent)?;
        h.follow_location(true)?;
        match self.method {
            "GET" => h.get(true)?,
            "PUT" => h.put(true)?,
            "POST" => h.post(true)?,
            other => h.custom_request(other)?,
        }

        let mut headers = List::new();
        headers.append(&format!("Host: {}", host))?;
        if data.len() > 0 {
            h.post_fields_copy(data)?;
            headers.append(&format!(
                "Accept: application/x-git-{}-result",
                self.service
            ))?;
            headers.append(&format!(
                "Content-Type: \
                 application/x-git-{}-request",
                self.service
            ))?;
        } else {
            headers.append("Accept: */*")?;
        }
        headers.append("Expect:")?;
        h.http_headers(headers)?;

        let mut content_type = None;
        let mut data = Vec::new();
        {
            let mut h = h.transfer();

            // Look for the Content-Type header
            h.header_function(|header| {
                let header = match str::from_utf8(header) {
                    Ok(s) => s,
                    Err(..) => return true,
                };
                let mut parts = header.splitn(2, ": ");
                let name = parts.next().unwrap();
                let value = match parts.next() {
                    Some(value) => value,
                    None => return true,
                };
                if name.eq_ignore_ascii_case("Content-Type") {
                    content_type = Some(value.trim().to_string());
                }

                true
            })?;

            // Collect the request's response in-memory
            h.write_function(|buf| {
                data.extend_from_slice(buf);
                Ok(buf.len())
            })?;

            // Send the request
            h.perform()?;
        }

        let code = h.response_code()?;
        if code != 200 {
            return Err(self.err(
                &format!(
                    "failed to receive HTTP 200 response: \
                     got {}",
                    code
                )[..],
            ));
        }

        // Check returned headers
        let expected = match self.method {
            "GET" => format!("application/x-git-{}-advertisement", self.service),
            _ => format!("application/x-git-{}-result", self.service),
        };
        match content_type {
            Some(ref content_type) if *content_type != expected => {
                return Err(self.err(
                    &format!(
                        "expected a Content-Type header \
                         with `{}` but found `{}`",
                        expected, content_type
                    )[..],
                ))
            }
            Some(..) => {}
            None => {
                return Err(self.err(
                    &format!(
                        "expected a Content-Type header \
                         with `{}` but didn't find one",
                        expected
                    )[..],
                ))
            }
        }

        // Ok, time to read off some data.
        let rdr = Cursor::new(data);
        self.reader = Some(rdr);

        // If there was a redirect, update the `CurlTransport` with the new base.
        if let Ok(Some(effective_url)) = h.effective_url() {
            let new_base = if effective_url.ends_with(self.url_path) {
                // Strip the action from the end.
                &effective_url[..effective_url.len() - self.url_path.len()]
            } else {
                // I'm not sure if this code path makes sense, but it's what
                // libgit does.
                effective_url
            };
            *self.base_url.lock().unwrap() = new_base.to_string();
        }

        Ok(())
    }
}

impl Read for CurlSubtransport {
    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
        if self.reader.is_none() {
            self.execute(&[])?;
        }
        self.reader.as_mut().unwrap().read(buf)
    }
}

impl Write for CurlSubtransport {
    fn write(&mut self, data: &[u8]) -> io::Result<usize> {
        if self.reader.is_none() {
            self.execute(data)?;
        }
        Ok(data.len())
    }
    fn flush(&mut self) -> io::Result<()> {
        Ok(())
    }
}