An attempt to build a Ruby module for a custom Ruby installation that has the
same major version as the system Ruby may unexpectedly cause the use of the
system Ruby library.
This closes#449 issue on GitHub.
The invocation parameters should be logged as well, notably the path of the file
that is failed to be created.
Also, log level changed to ALERT as it's quite critical error.
Earlier, if nxt_mp_create() failed to allocate memory while accepting a new
connection, the resulting NULL was subsequently passed to nxt_mp_destroy(),
crashing the process.
More, if nxt_mp_create() was successful but nxt_sockaddr_cache_alloc() failed,
the connection object wasn't destroyed properly, leaving the connection counter
in an inconsistent state. Repeated, this condition lowered the connection
capacity of the process and could eventually prevent it from accepting
connections altogether.
Previously, the log message callback used a generic log function, that relied on the process time cache.
Since there were no time update calls in the application processes, all log lines were printed with the
same time, usually correlated with the process start.
Now, a non-cached logging function from libunit is used.
This makes log format used in libunit consistent with the daemon, where milliseconds are printed only in the
debug log level.
Currently a compile time switch is used, since there's no support for runtime changing of a log level for now.
But in the future this should be a runtime condition, similar to nxt_log_time_handler().
According to the C standard, pointer arguments passed to memcpy() calls shall
still have valid values. NULL is considered as invalid.
Found with GCC Static Analyzer.
This is useful to escape "/" in path fragments. For example, in order
to reference the application named "foo/bar":
{
"pass": "applications/foo%2Fbar"
}
One of the ways to detect Unit's startup and subsequent readiness to accept
commands relies on waiting for the control socket file to be created.
Earlier, it was unreliable due to a race condition between the client's
connect() and the daemon's listen() calls after the socket's bind() call.
Now, unix domain listening sockets are created with a nxt_listen_socket_create()
call as follows:
s = socket();
unlink("path/to/socket.tmp")
bind(s, "path/to/socket.tmp");
listen(s);
rename("path/to/socket.tmp", "path/to/socket");
This eliminates a time-lapse when the socket file is already created but nobody
is listening on it yet, which therefore prevents the condition described above.
Also, it allows reliably detecting whether the socket is being used or simply
wasn't cleaned after the daemon stopped abruptly. A successful connection to
the socket file means the daemon has been started; otherwise, the file can be
overwritten.
The "return" action can be used to immediately generate a simple HTTP response
with an arbitrary status:
{
"action": {
"return": 404
}
}
This is especially useful for denying access to specific resources.
It allows proceeding to another action if a file isn't available.
An example:
{
"share": "/data/www/",
"fallback": {
"pass": "applications/php"
}
}
In the example above, an attempt is made first to serve a request with
a file from the "/data/www/" directory. If there's no such file, the
request is passed to the "php" application.
Fallback actions may be nested:
{
"share": "/data/www/",
"fallback": {
"share": "/data/cache/",
"fallback": {
"proxy": "http://127.0.0.1:9000"
}
}
}
A check for the ".php" extension is added to prevent execution of files
with arbitrary extensions in cases where "index" and "script" options
aren't used.
A quote from the Python 3 documentation:
| When interactive, stdout and stderr streams are line-buffered.
| Otherwise, they are block-buffered like regular text files.
As a result, if an exception occurred and PyErr_Print() was called, its output
could be buffered but not printed to the log for a while (ultimately, until
the interpreter finalization). If the application process crashed shortly,
the backtrace was completely lost.
Buffering can be disabled by redefining the sys.stderr stream object.
However, interference with standard environment objects was deemed undesirable.
Instead, sys.stderr.flush() is called every time after printing exceptions.
A potential advantage here is that lines from backtraces won't be mixed
with other lines in the log.
PyObject_HasAttrString() is just a wrapper over PyObject_GetAttrString(),
while PyObject_CallMethod() calls it as the first step. As a result,
PyObject_GetAttrString() was called twice if close() was present.
To get rid of PyObject_HasAttrString() while keeping the same behaviour,
the PyObject_CallMethod() call has been decomposed into separate calls of
PyObject_GetAttrString() and PyObject_CallFunction().
On success, PyObject_CallMethod() returns a new reference to
the result of the call, which previously got lost.
Also, error logging on failure was added.
The issue was introduced by b0148ec28c4d.
According to the documentation, PyObject_GetIter():
| Raises TypeError and returns NULL if the object cannot be iterated.
Previously, this exception wasn't printed or cleared and remained unhandled.
According to the documentation, PyIter_Next():
| If there are no remaining values, returns NULL with no exception set.
| If an error occurs while retrieving the item, returns NULL and passes
| along the exception.
Previously, this exception wasn't properly handled and the response was
finalized as successful.
This issue was introduced in b0148ec28c4d.
A check for PyErr_Occurred() located in the code below might print this
traceback or occasionally catch an exception from one of the two response
close() calls.
Albeit that exceptions from the close() calls also need to be catched,
it's clear that this particular check wasn't supposed to do so. This is
another issue and it will be fixed later.
This field was intended for MIME type lookup by file extension when serving
static files, but this use case is too narrow; only a fraction of requests
targets static content, and the URI presumably isn't rewritten. Moreover,
current implementation uses the entire filename for MIME type lookup if the
file has no extension.
Instead of extracting filenames and extensions when parsing requests, it's
easier to obtain them right before serving static content; this behavior is
already implemented. Thus, we can drop excessive logic from parser.
Earlier, the paths were normalized only if there was a "/" at the end, which
is wrong according to section 5.2.4 of RFC 3986 and hypothetically may allow
to the directory above the document root.
In theory, all space characters in request target must be encoded; however,
some clients may violate the specification. For the sake of interoperability,
Unit supports unencoded space characters.
Previously, if there was a space character before the extension or arguments
parts, those parts weren't recognized. Also, quoted symbols and complex
target weren't detected after a space character.
Valentin Bartenev