A glob pattern is a text expression that matches one or more
file names using wild cards familiar to most users of a command line.
* is a glob that matches any name at all and
Readme.txt is a glob that matches exactly one file. For purposes of
Fossil's globs, a file name with a directory prefix is "just a string"
and the globs do not apply any special meaning to the directory part
of the name. Thus the glob
* matches any name, including any
directory prefix, and
*/* matches a name with one or more
A glob should not be confused with a regular expression (RE), even though they use some of the same special characters for similar purposes, because they are not fully compatible pattern matching languages. Fossil uses globs when matching file names with the settings described in this document, not REs.
These settings hold one or more file glob patterns to cause Fossil to give matching named files special treatment. Glob patterns are also accepted in options to certain commands and as query parameters to certain Fossil UI web pages.
Where Fossil also accepts globs in commands, this handling may interact with your OS’s command shell or its C runtime system, because they may have their own glob pattern handling. We will detail such interactions below.
Where Fossil accepts glob patterns, it will usually accept a list of such patterns, each individual pattern separated from the others by white space or commas. If a glob must contain white spaces or commas, it can be quoted with either single or double quotation marks. A list is said to match if any one glob in the list matches.
A glob pattern matches a given file name if it successfully consumes and matches the entire name. Partial matches are failed matches.
Most characters in a glob pattern consume a single character of the file name and must match it exactly. For instance, “a” in a glob simply matches the letter “a” in the file name unless it is inside a special character sequence.
Other characters have special meaning, and they may include otherwise normal characters to give them special meaning:
||Matches any sequence of zero or more characters|
||Matches exactly one character|
||Matches one character from the enclosed list of characters|
||Matches one character not in the enclosed list|
Note that unlike POSIX globs, these special characters and
sequences are allowed to match
/ directory separators as well as the
. in the name of a hidden file or directory. This is because
Fossil file names are stored as complete path names. The distinction
between file name and directory name is “below” Fossil in this sense.
The bracket expressions above require some additional explanation:
A range of characters may be specified with
[a-f]matches exactly the same characters as
[abcdef]. Ranges reflect Unicode code points without any locale-specific collation sequence. Therefore, this particular sequence never matches the Unicode pre-composed character
é, for example. (U+00E9)
This dependence on character/code point ordering may have other effects to surprise you. For example, the glob
[A-z]not only matches upper and lowercase ASCII letters, it also matches several punctuation characters placed between
ain both ASCII and Unicode:
_, and `.
You may include a literal
-in a list by placing it last, just before the
You may include a literal
]in a list by making the first character after the
[^. At any other place,
]ends the list.
You may include a literal
^in a list by placing it anywhere except after the opening
Beware that a range must be specified from low value to high value:
[z-a]does not match any character at all, preventing the entire glob from matching.
Some examples of character lists:
||Matches any one of
||Matches exactly one character other than
||Matches exactly one hexadecimal digit|
||Matches exactly one character other than
||Matches exactly one character other than
White space means the specific ASCII characters TAB, LF, VT, FF, CR, and SPACE. Note that this does not include any of the many additional spacing characters available in Unicode such as U+00A0, NO-BREAK SPACE.
Because both LF and CR are white space and leading and trailing spaces are stripped from each glob in a list, a list of globs may be broken into lines between globs when the list is stored in a file, as for a versioned setting.
Note that 'single quotes' and "double quotes" are the ASCII straight quote characters, not any of the other quotation marks provided in Unicode and specifically not the "curly" quotes preferred by typesetters and word processors.
File Names to Match
Before it is compared to a glob pattern, each file name is transformed to a canonical form:
- all directory separators are changed to
- redundant slashes are removed
.path components are removed
..path components are resolved
(There are additional details we are ignoring here, but they cover rare edge cases and follow the principle of least surprise.)
The glob must match the entire canonical file name to be considered a match.
The goal is to have a name that is the simplest possible for each particular file, and that will be the same regardless of the platform you run Fossil on. This is important when you have a repository cloned from multiple platforms and have globs in versioned settings: you want those settings to be interpreted the same way everywhere.
Beware, however, that all glob matching in Fossil is case sensitive
regardless of host platform and file system. This will not be a surprise
on POSIX platforms where file names are usually treated case
sensitively. However, most Windows file systems are case preserving but
case insensitive. That is, on Windows, the names
are usually names of the same file. The same is true in other cases,
such as by default on macOS file systems and in the file system drivers
for Windows file systems running on non-Windows systems. (e.g. exfat on
Linux.) Therefore, write your Fossil glob patterns to match the name of
the file as checked into the repository.
Some example cases:
||Matches only a file named
||Matches every C source or header file in the tree at the root or at any depth. Again, this is (deliberately) different from Unix file globs and Windows wild cards.|
Where Globs are Used
Settings that are Globs
These settings are all lists of glob patterns:
||Files that should be treated as binary files for committing and merging purposes|
||Files that the
||Files in which it is okay to have
||Alias for the
||Files that the
||Files that the
||Files that the
All may be versioned, local, or global. Use
settings to manage local and global settings, or a file in the
.fossil-settings/ folder at the root of the tree named
for each for versioned setting.
Using versioned settings for these not only has the advantage that they are tracked in the repository just like the rest of your project, but you can more easily keep longer lists of more complicated glob patterns than would be practical in either local or global settings.
ignore-glob is an example of one setting that frequently grows
to be an elaborate list of files that should be ignored by most
commands. This is especially true when one (or more) IDEs are used in
a project because each IDE has its own ideas of how and where to cache
information that speeds up its browsing and building tasks but which
need not be preserved in your project's history.
Commands that Refer to Globs
Many of the commands that respect the settings containing globs have
options to override some or all of the settings. These options are
usually named to correspond to the setting they override, such as
--ignore to override the
ignore-glob setting. These commands are:
zip produce compressed archives of a
specific checkin. They may be further restricted by options that
specify glob patterns that name files to include or exclude rather
than archiving the entire checkin.
implement or support with web servers provide a mechanism to name some
files to serve with static content where a list of glob patterns
specifies what content may be served.
Web Pages that Refer to Globs
/timeline page supports the query parameter
names a list of glob patterns defining which files to focus the
timeline on. It also has the query parameters
names a tag to focus on, which can be configured with
use a glob pattern to match tag names instead of the default exact
match or a couple of other comparison styles.
/zip generate compressed archives
of a specific checkin. They may be further restricted by query
parameters that specify glob patterns that name files to include or
exclude rather than taking the entire checkin.
Fossil glob patterns are based on the glob pattern feature of POSIX
shells. Fossil glob patterns also have a quoting mechanism, discussed
above. Because other parts of your operating system may interpret glob
patterns and quotes separately from Fossil, it is often difficult to
give glob patterns correctly to Fossil on the command line. Quotes and
special characters in glob patterns are likely to be interpreted when
given as part of a
fossil command, causing unexpected behavior.
These problems do not affect versioned settings files
or Admin → Settings in Fossil UI. Consequently, it is better to
*-glob settings via these methods than to use
That advice does not help you when you are giving one-off glob patterns
fossil commands. The remainder of this section gives remedies and
workarounds for these problems.
If you are using Fossil on a system with a POSIX-compatible shell
— Linux, macOS, the BSDs, Unix, Cygwin, WSL etc. — the shell
may expand the glob patterns before passing the result to the
Sometimes this is exactly what you want. Consider this command for example:
$ fossil add RE*
If you give that command in a directory containing
RELEASE-NOTES.txt, the shell will expand the command to:
$ fossil add README.txt RELEASE-NOTES.txt
…which is compatible with the
fossil add command's argument list,
which allows multiple files.
Now consider what happens instead if you say:
$ fossil add --ignore RE* src/*.c
This does not do what you want because the shell will expand both
src/*.c, causing one of the two files matching the
pattern to be ignored and the other to be added to the repository. You
need to say this in that case:
$ fossil add --ignore 'RE*' src/*.c
The single quotes force a POSIX shell to pass the
RE* glob pattern
through to Fossil untouched, which will do its own glob pattern
matching. There are other methods of quoting a glob pattern or escaping
its special characters; see your shell's manual.
Beware that Fossil's
--ignore option does not override explicit file
$ fossil add --ignore 'REALLY SECRET STUFF.txt' RE*
You might think that would add everything beginning with
REALLY SECRET STUFF.txt, but when a file is both given
explicitly to Fossil and also matches an ignore rule, Fossil asks what
you want to do with it in the default case; and it does not even ask
if you gave the
--force option along with
The spaces in the ignored file name above bring us to another point: such file names must be quoted in Fossil glob patterns, lest Fossil interpret it as multiple glob patterns, but the shell interprets quotation marks itself.
One way to fix both this and the previous problem is:
$ fossil add --ignore "'REALLY SECRET STUFF.txt'" READ*
The nested quotation marks cause the inner set to be passed through to
Fossil, and the more specific glob pattern at the end — that is,
RE* — avoids a conflict between explicitly-listed
--ignore rules in the
fossil add command.
Another solution would be to use shell escaping instead of nested quoting:
$ fossil add --ignore "\"REALLY SECRET STUFF.txt\"" READ*
It bears repeating that the two glob patterns here are not interpreted
the same way when running this command from a subdirectory of the top
checkout directory as when running it at the top of the checkout tree.
If these files were in a subdirectory of the checkout tree called
and that was your current working directory, the command would have to
$ fossil add --ignore "'doc/REALLY SECRET STUFF.txt'" READ*
instead. The Fossil glob pattern still needs the
doc/ prefix because
Fossil always interprets glob patterns from the base of the checkout
directory, not from the current working directory as POSIX shells do.
When in doubt, use
fossil status after running commands like the
above to make sure the right set of files were scheduled for insertion
into the repository before checking the changes in. You never want to
accidentally check something like a password, an API key, or the
private half of a public cryptographic key into Fossil repository that
can be read by people who should not have such secrets.
Before we get into Windows-specific details here, beware that this section does not apply to the several Microsoft Windows extensions that provide POSIX semantics to Windows, for which you want to use the advice in the POSIX section above instead:
- the ancient and rarely-used Microsoft POSIX subsystem;
- its now-discontinued replacement feature, Services for Unix; or
- their modern replacement, the Windows Subsystem for Linux
(The latter is sometimes incorrectly called "Bash on Windows" or "Ubuntu on Windows," but the feature provides much more than just Bash or Ubuntu for Windows.)
Neither standard Windows command shell —
cmd.exe or PowerShell
— expands glob patterns the way POSIX shells do. Windows command
shells rely on the command itself to do the glob pattern expansion. The
way this works depends on several factors:
- the version of Windows you are using
- which OS upgrades have been applied to it
- the compiler that built your Fossil executable
- whether you are running the command interactively
- whether the command is built against a runtime system that does this at all
- whether the Fossil command is being run from a file named
*.BATvs being named
Usually (but not always!) the C runtime library that your
executable is built against does this glob expansion on Windows so the
program proper does not have to. This may then interact with the way the
Windows command shell you’re using handles argument quoting. Because of
these differences, it is common to find perfectly valid Fossil command
examples that were written and tested on a POSIX system which then fail
when tried on Windows.
The most common problem is figuring out how to get a glob pattern passed
on the command line into
fossil.exe without it being expanded by the C
runtime library that your particular Fossil executable is linked to,
which tries to act like the POSIX systems described above. Windows is
not strongly governed by POSIX, so it has not historically hewed closely
to its strictures.
For example, consider how you would set
* in order to
get normal Windows text files with CR+LF line endings past Fossil's
"looks like a binary file" check. The naïve approach will not work:
C:\...> fossil setting crlf-glob *
The C runtime library will expand that to the list of all files in the
current directory, which will probably cause a Fossil error because
Fossil expects either nothing or option flags after the setting's new
value, not a list of file names. (To be fair, the same thing will happen
on POSIX systems, only at the shell level, before
gets run by the shell.)
Let's try again:
C:\...> fossil setting crlf-glob '*'
Quoting the argument like that will work reliably on POSIX, but it may
or may not work on Windows. If your Windows command shell interprets the
quotes, it means
fossil.exe will see only the bare
* so the C
runtime library it is linked to will likely expand the list of files in
the current directory before the
setting command gets a chance to
parse the command line arguments, causing the same failure as above.
This alternative only works if you’re using a Windows command shell that
passes the quotes through to the executable and you have linked Fossil
to a C runtime library that interprets the quotes properly itself,
resulting in a bare
* getting clear down to Fossil’s
An approach that will work reliably is:
C:\...> echo * | fossil setting crlf-glob --args -
This works because the built-in Windows command
echo does not expand its
arguments, and the
--args - option makes Fossil read further command
arguments from its standard input, which is connected to the output
echo by the pipe. (
- is a common Unix convention meaning
"standard input," which Fossil obeys.) A batch script to automate this trick was
posted on the now-inactive Fossil Mailing List.
(Ironically, this method will not work on POSIX systems because it is
not up to the command to expand globs. The shell will expand the
echo command, so the list of file names will be passed to the
fossil standard input, just as with the first example above!)
Another (usually) correct approach which will work on both Windows and POSIX systems:
C:\...> fossil setting crlf-glob *,
This works because the trailing comma prevents the glob pattern from
matching any files, unless you happen to have files named with a
trailing comma in the current directory. If the pattern matches no
files, it is passed into Fossil's
main() function as-is by the C
runtime system. Since Fossil uses commas to separate multiple glob
patterns, this means "all files from the root of the Fossil checkout
directory downward and nothing else," which is of course equivalent to
"all managed files in this repository," our original goal.
To preview the effects of command line glob pattern expansion for
various glob patterns (unquoted, quoted, comma-terminated), for any
combination of command shell, OS, C run time, and Fossil version,
precede the command you want to test with
test-echo like so:
$ fossil test-echo setting crlf-glob "*" C:\> echo * | fossil test-echo setting crlf-glob --args -
test-glob command is also handy to test if a string
matches a glob pattern.
Many other version control systems handle the specific case of ignoring certain files differently from Fossil: they have you create individual "ignore" files in each folder, which specify things ignored in that folder and below. Usually some form of glob patterns are used in those files, but the details differ from Fossil.
In many simple cases, you can just store a top level "ignore" file in
.fossil-settings/ignore-glob. But as usual, there will be lots of
Git has a rich collection of ignore files which accumulate rules that affect the current command. There are global files, per-user files, per workspace unmanaged files, and fully version controlled files. Some of the files used have no set name, but are called out in configuration files.
In contrast, Fossil has a global setting and a local setting, but the local setting
overrides the global rather than extending it. Similarly, a Fossil
--ignore option replaces the
ignore-glob setting rather
than extending it.
With that in mind, translating a
.gitignore file into
.fossil-settings/ignore-glob may be possible in many cases. Here are
some of features of
.gitignore and comments on how they relate to
- "A blank line matches no files...": same in Fossil.
- "A line starting with # serves as a comment....": not in Fossil.
- "Trailing spaces are ignored unless they are quoted..." is similar in Fossil. All whitespace before and after a glob is trimmed in Fossil unless quoted with single or double quotes. Git uses backslash quoting instead, which Fossil does not.
- "An optional prefix "!" which negates the pattern...": not in Fossil.
- Git's globs are relative to the location of the
.gitignorefile: Fossil's globs are relative to the root of the workspace.
- Git's globs and Fossil's globs treat directory separators differently. Git includes a notation for zero or more directories that is not needed in Fossil.
In a project with source and documentation:
work +-- doc +-- src
doc/.gitignore might contain:
# Finished documents by pandoc via LaTeX *.pdf # Intermediate files *.tex *.toc *.log *.out *.tmp
.fossil-settings/ignore-glob with similar effect, also
limited to the
doc/*.pdf doc/*.tex, doc/*.toc, doc/*.log, doc/*.out, doc/*.tmp
Implementation and References
The implementation of the Fossil-specific glob pattern handling is here:
||pattern list loading, parsing, and generic matching code|
||application of glob patterns to file names|
See the Adding Features to Fossil document for broader details about finding and working with such code.
The actual pattern matching leverages the
GLOB operator in SQLite, so
you may find its documentation, source code and test