pseudolog man page
pseudolog — pseudo log parser
pseudolog -l [-Pv] [ -E timeformat ] [ -x flags ] [Specifications]
pseudolog [-UPv] [ -E timeformat ] [ -F format ] [ -x flags ]
pseudolog -D [-Pv] [ -E timeformat ] [ -x flags ] [Specifications] [Specifications]
The pseudolog utility displays, creates, or deletes log entries associated with the pseudo daemon. Creation of log entries is useful only to create timestamps or notes; for instance, you could create a log entry before beginning a process, so there would be a timestamp for the beginning of that process. There are a number of special options used to match or create the components of a log entry; these are called specifications, and are detailed in the Specifications section below.
The following other options are supported:
Print a usage message and exit.
Delete rows selected by the query. This is not reversible.
- -E timeformat
Specify a format string (for strptime(3) or strftime(3) to use) for displaying or interpreting time stamps. The same format is used both for parsing and displaying stamps.
- -F format
Specifies a format string for displaying log entries. This format cannot be used to create log entries, only for display. The format string is a printf(3) type format string, with format specifiers matching the option characters used in specifications (see Specifications). There are some limitations on allowed formats, and misuse of this feature could cause interesting or surprising failures.
Create a log entry. This option is mutually exclusive with the -F option, or with any relative specifications (see below).
- -P path
Specify that path should be used as the PSEUDO_PREFIX value, overriding any environment setting.
Restrict query output to unique rows. Rows will have members defined by the -F (format) option. If all members are the same between two rows, only one is displayed. Applies only to queries.
Increase verbosity (debug level). Not useful except when debugging pseudo. Deprecated; use -x.
Specify debugging flags of interest. Not useful except when debugging pseudo.
Other option characters are defined as specifications, and all of those require arguments to specify their values.
The various components of a log entry can be specified, either as command-line options, or as format specifiers. In either case, the same character is used for a given component of a log entry. When querying values, one of the following prefixes may be prepended to a value; otherwise, the value is used for a literal match (an SQL = operator).
Greater than; true if the related field is greater than the provided value.
Less than; true if the related field is less than the provided value.
Bitwise and; true if the related field, bitwise-and the provided value, is non-zero. (This is useful primarily for permissions or modes.)
Equal to. (This is a no-op, as of this writing.)
Not equal to.
Similar to ~. This is valid only on text fields, and is equivalent to the SQL LIKE operator, with % patterns on the ends; it performs an unanchored, case-insensitive match.
Similar to %. This is valid only on text fields, and is equivalent to the SQL LIKE operator, but performs an anchored match. The match is case-insensitive. The specifier ~%foo% is equivalent to the specifier %foo.
Unlike. This is the inverse of ~; it specifies NOT LIKE.
Escape the string. This is useful if you want to have one of the other modifiers at the beginning of the string.
Only =and\ modifiers may be used in conjunction with the -l option.
The following characters correspond to specific fields in a log entry. In general, numeric values are parsed in the standard C idiom (where a leading 0 indicates an octal value, and a leading 0x indicates a hexadecimal value, and any other number is decimal). A few fields are parsed or displayed in other ways, as detailed in their entries.
Access mode. This is an access mode specified in the form used by fopen(3), such as "r+" to indicate read/write access. Note that specifying a as an access mode will include non-append writes, as the "a" mode implies write and append both. This feature is slightly experimental and may not correctly identify the access type of every access. The string x may be specified to indicate execute access.
Client ID (the PID of a client).
Device number (from a stat buffer).
File descriptor. In some cases, messages have an associated file descriptor identified.
GID. The group ID associated with an entry.
Tag. This is a text field. In log entries created by pseudo, this field holds the value that the environment variable PSEUDO_TAG had in the client's environment.
Inode number (from a stat buffer).
ID. This is the database row number. Normally these are assigned as monotonically increasing values as rows are inserted, making them a more reliable sorting mechanism than timestamps. The default ordering is by ID. m Permissions. These can be entered as an octal value or as a symbolic mode string, similar to the output of ls(1) -l. The file type component is ignored.
Mode. This can be entered as an octal value or as a symbolic mode string, similar to the output of ls(1) -l. This is tested against the whole file mode, including both the type and permissions bits. In general, it is more useful to use the m or t specifiers.
Operation. This is the name of the file system operation (e.g., "open" or "rename").
Order. This takes another specification character as the field on which to order results. A '<' implies a descending order sort, a '>' or no modifier specifies an ascending order sort. By default, records are sorted by ID.
File path. This is a text field.
Result. This is the pseudo result code, most often "fail" or "succeed". Note that "fail" doesn't mean that an underlying operation failed; for instance, if a "stat" operation fails, it usually means that there was no entry in the pseudo database.
Program. This is the program name (as retrieved by glibc's program_invocation_name variable), which has the full path if and only if the program was invoked by full path name.
Timestamp. The format of this field is controlled by the -E format string, which is used with strftime(3) when displaying entries, or with strptime(3) when interpreting command line values. There is a small selection of common default time formats understood by the parser. Time fields not specified default to the current time. Note that specifying a time stamp when creating a log entry may yield confusing results.
Severity. Log messages can have a severity, with the default for file operations being "info". t File type. This corresponds to the first letter of a mode string, or the values accepted by the -type option to find(1). This is compared only against the file type bits of a mode.
Text. This is an optional field available for user use when creating log entries, or to hold the text of an error message when an error is logged. It is, of course, a text field.
UID. The user ID associated with an entry.
Type. This is usually "op" for operations, or "ping" for the ping messages clients send to confirm server availability. Other types should rarely occur, but include "ack" and "nak" for server responses (which are never logged), and "halt" for shutdown messages (currently not logged).
The following examples illustrate some of the likely usage patterns for pseudolog.
- pseudolog -m '&020' -t d
Report on all directories which are group-writeable.
- pseudolog -m 755 -t f
Report on all plain files which have the mode rwxr-xr-x.
- pseudolog -s '>03:19:00' -s '<03:20:00'
Report on all entries created after 03:19:00 and before 03:20:00 on the current date.
- pseudolog -p '~/usr/bin/%' -F '%-8o %p'
Report on every entry with a path beginning with the string '/usr/bin', displaying the operation name (in a space-padded field of eight characters, left-adjusted) followed by the path.
- pseudolog -l -T 'stamp test'
Create an entry with all fields zero or blank, except for the text field, which is set to the text "stamp test", and the timestamp, which is set to the current time.
- pseudolog -D -r succeed -F '%p' -O p
Display all paths for which operations succeeded, sorted by path value.
The only environment variable supported by pseudolog is:
If set, the variable PSEUDO_PREFIX is used to determine the path to use to find the logs.db database file, in PSEUDO_PREFIX/var/pseudo.
The user might think our intent is to replace all of SQL. It's not. If the options here aren't enough, rather than adding more options to this already fairly elaborate program, just do raw SQL queries on the logs.db file.
The formatting options are handled by converting them into printf(3) format strings, without much checking. As a result, it is possible for a malformed format string to cause printf() to explode unexpectedly.