1 # Configuration of the UCW library and related utilities
3 ######## Memory Mapped Access to Files ##########################################
5 # Whenever you specify 0 for I/O buffer size, memory mapping is used instead.
8 # Map this many bytes at once (needs to be a multiple of CPU page size)
11 # When in need to extend a file, grow it by so many bytes (>= page size)
16 ######## Direct Streamed I/O on Files ###########################################
20 # Debug: Cheat by turning off O_DIRECT
25 ######## Atomic Multi-Threaded I/O on Files #####################################
34 ######## Parametrized I/O on Files ##############################################
40 # Access type (std|direct|mmap).
43 # Size of I/O buffer. Something of the order of megabytes for fast disks is recommended for direct I/O.
46 # Optimize for mixed forward/backward reading (standard I/O only)
49 # Perform read-ahead (direct I/O only)
52 # Maximum number of write-back requests queued (direct I/O only)
59 ######## Temporary files ########################################################
63 # By default, we use the system's default temporary directory ($TMPDIR or /tmp),
64 # but sometimes it is better to store the temporary files in the local tree.
65 Dir @INSTALL_VAR_PREFIX@tmp
67 # Prefix of temporary file names
70 # By default, we append a random number to Prefix to get a temporary file name.
71 # If Prefix points to a directory that is not writable by malicious users,
72 # we can be less careful and use more consistent names of temporary files
73 # formed by adding "pid(-tid)-counter" instead.
78 ######## Threads ################################################################
82 # Default thread stack size
87 ######## Sorter #################################################################
91 # Trace sorting (1=basic statistics, 2=more stats, 3 and more for debugging)
94 # Trace array sorting (internal sorters)
97 # How much memory is the sorter allowed to use
100 # File access used by the sorter (see FBParam section for details)
103 # Use a different file access method for small inputs (less than the specified size)
104 SmallFileAccess std 64K
107 # Min-/Maximum number of bits to use in the external radix-sort (beware, we will open
108 # 1+2^this files and require a stream buffer for each of them; however, while we are
109 # doing that, the sort buffer is not allocated). Set both to zero to disable radix-sorting.
113 # The same for multi-way merging. The memory requirements are also the same,
114 # but please keep in mind that this can create lots of SortBuffer-sized files,
115 # so it is probably better to keep it disabled if you have a small SortBuffer.
119 # If we did not use radix-sorter to the full width, we still might add some more
120 # bits to the width to get chunks which are even smaller than SortBuffer, because
121 # it can speed up internal sorting later. However, we also want to avoid small
122 # files, so we add only a little.
125 # Number of threads used for sorting (0=disable threading)
128 # Minimum size of input (in bytes) to consider multi-threaded internal sorting
131 # Chunks smaller than ThreadThreshold are sorted by a sequential algorithm, but
132 # if they are at least of the following size, different chunks are sorted in
133 # parallel. There is a slight space penalty for setting up the parallel process,
134 # so better avoid setting this number too small.
137 # Internal radix-sort stops at this size and switches to QuickSort (must be >0)
140 # Debugging switches (see the source)
145 ######## URL processing #########################################################
149 # Ignore spaces at the start/end of a URL
152 # Ignore underflows in relative paths (/../ from root)
155 # Some URL's with many repeated components are filtered out to avoid infinite
156 # URL's (e.g. http://czech.recoder.cz/win/iso/win/iso/file.html, or
157 # http://a.com/?a=b&a=b&a=b, ...).
158 # The URL is split to components divided by any of the specified separators.
159 # Then the separators are forgotten and the components between them are
161 ComponentSeparators /&?
163 # URL is filtered out if there's a sequence of components in a row with at most
164 # MaxRepeatLength components and the sequence is repeated more than MinRepeatCount
165 # times. Default values are high MinRepeatCount and low MaxRepeatLength, so the
166 # mechanism is disabled.
170 # Maximum number of occurences of a single component in the entire URL (possibly interleaved
171 # by different components). The detector is disabled by default.
176 ######## Logging ################################################################
180 # In this section, you can define various logging streams which can be referred to by other sections.
183 # # The name of the stream
186 # # When it should log the messages to a file, a name of the file should be specified.
187 # # Escape sequences for current date and time as described in strftime(3) can be used.
188 # FileName log/test-%Y%m%d
190 # # If you need to log to stderr or another already opened descriptor, you can specify its number.
193 # # Instead of a file, a syslog facility can be specified. See syslog(3) for an explanation.
194 # SyslogFacility daemon
196 # # You can request that syslog includes a process ID in each message. Due to inflexibility
197 # # of the syslog protocol, all syslog streams active at a moment must agree on this setting.
201 # # When logging to files, timestamps with microsecond precision can be requested. (default: 0)
204 # # Messages logged to this stream can be restricted to a subset of severity levels.
205 # # Available levels are: debug info warn error info_r warn_r error_r fatal.
206 # # This configuration item is a bitmap with a default of "all", so we need the ":reset" operator.
207 # Levels:reset info warn error fatal
209 # # Similarly, messages can be restricted to a subset of message types. The types are
210 # # specific for each program. This configuration item is a list of type names; by default
211 # # it is empty, which is equivalent to all types being enabled.
212 # Types:reset default foo
214 # # Should the message types be logged? They usually do not carry much useful
215 # # information for the viewer of the log, so they are not included by default,
216 # # but you might want to see them when tuning the Types setting. (default: 0)
219 # # If an error occurs when logging a message to this stream, the program normally
220 # # logs a special error message to the other streams and continues running. You can
221 # # however request to exit the program in such cases, so that the log files are
222 # # guaranteed to be complete. (default: 0)
225 # # Let stderr of the program point to this file-based log_stream (default: 0)
228 # # Some events are logworthy, but they could happen too frequently and flood the log.
229 # # You can avoid the flooding by setting up a rate limiter for a specific subset of
230 # # message types. If more limiters match the type of a message, only the last one applies.
232 # # A list of message types (default: empty = all types)
235 # # The maximum allowed sustained rate (messages/second, may be fractional)
238 # # Maximum length of a burst temporarily exceeding the rate (default: try to guess)
242 # # The messages that have passed the filters and limiters can be forwarded to other
243 # # log streams. Logging loops are not healthy for your program :) (a list of stream names)
244 # Substream another-stream