These were only used to construct constant strings with additional
length fields. However, we can just use standard constant character
arrays instead and let the compiler calculate the string length.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
strncmp() isn't intended to be a secure strcmp() replacement, it is
designed to be used if you want to compare the first n characters of two
strings. Since we always compare character pointers with string
literals, switch to using strcmp() everywhere.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
There's no need to reset windows unless the layout was changed. Move
wins_reset() to a separate branch to reflect this.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Remove some redundant wnoutrefresh() invocations. There's no need to
copy a window to the virtual screen unless doupdate() is invoked
immediately afterwards.
This reduces flicker when browsing in the calendar panel.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
This was the only big block of code left in our main loop. Move it to a
separate function, making the main loop a bit clearer.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Refactor the logic inside keys_display_bindings_bar() and remove the
need to place the "show next page" key binding at the right positions.
This used to be a pain to maintain, since we always had to move key
bindings around when introducing a new key. Fix this by passing the
actual key bindings in an array and using a separate parameter for the
"show next page" key binding (which is automatically inserted at the
right places from now on).
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We don't change any configuration nor do we modify items here, so we can
safely invoke calcurse in read-only mode. Fixes a couple of permission
errors seen with `make distcheck`.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We don't save any configuration nor items if this is set. This should be
used with care, and hence there's no short option for this.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We really don't want to use the calcurse(1) binary from "$PATH" here.
Introduce an environment variable "$CALCURSE" which can be used specify
the binary to use and set this variable to the binary located in the
build directory by default. Similarly, add another environment variable
"$DATA_DIR" that can be used to specify a data directory.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Add tests and the dummy data directory to "EXTRA_DIST" since we want
these to be shipped in distribution tarballs. Spotted with `make
distcheck`.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Make commenting style consistent with everywhere else. Use "/* [...] */"
instead of "// [...]".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Now that we're fully POSIX compliant and fixed up some minor issues with
ISO C99, use "-std=c99 -pedantic -D_POSIX_C_SOURCE=200809L" as default
build flags for calcurse and the test suite.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
All members of the tm structure are signed integers. Stick to using the
correct fscanf() format specifiers for these.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
ISO C99 forbids unnamed structs/unions and nested functions. Move the
"HTABLE_HEAD" and "HTABLE_GENERATE" parts out of the function body and
place them at the very top of the file (where function definitions are
allowed). Also, remove the unnamed struct from "htable.h" (which was
useless anyway).
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
* "WAIT_MYPGRP" isn't POSIX'ish. Relying on this caused compilation
issues in certain environments (e.g. under Cygwin). As a workaround,
define "WAIT_MYPGRP" explicitly if it's undefined.
* "P_tmpdir" is an XSI extension. Don't try use it if it isn't
available.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
The bzero() and bcopy() functions are deprecated and were removed from
the POSIX standard in IEEE Std. 1003.1-2008. Remove all usages of
bzero()/bcopy() and replace them by appropriate memset()/memmove()
calls.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
The last argument to strncat() should not be the total buffer length; it
should be the space remaining:
The strncat() function shall append not more than n bytes (a null
byte and bytes that follow it are not appended) from the array
pointed to by s2 to the end of the string pointed to by s1. The
initial byte of s2 overwrites the null byte at the end of s1. A
terminating null byte is always appended to the result.
This patch fixes a couple of potential buffer overflow vulnerabilities.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Speeds up execution, prevents failures when pager or log file contains
spaces, fixes a warning seen with "-Wunused-result".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Skip the newline check if fgets() returns a NULL string. Fixes another
warning seen with "-Wunused-result".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Be more restrictive with what we allow in data files and bail out if a
line doesn't conform to our specification, especially in regard to
separators. This prevents unexpected behavior when a data file is edited
manually. As a bonus, this fixes a whole pile of compiler warnings
previously seen with "-Wunused-result".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Ensure we don't read arbitrary data when fgets() returns a NULL string
(meaning that either the EOF is encountered or an error occurred). This
also fixes a couple of compiler warnings seen with "-Wunused-result".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Append missing configuration variables to the end of our configuration
file instead of keeping them undefined.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Reintroduce a map for configuration variable parsing. We use a different
approach this time: Each map entry contains
* a key,
* a callback that can be used to parse that variable,
* a callback that can be used to serialize that variable and
* a target buffer that the parsed value is written to/read from.
Commits 4f4891bdb88410ae04225f3d6acfa31d73a3901a and
6377582841118688aee13aff98c9216403582e45 show that we are pretty
undecided on using a map or not. However, now that we use parser and
serialization wrappers for every variable, having a central map makes
everything much cleaner. The runtimes of config_load() and config_save()
are slightly increased (by a constant factor). This will also allow us
for implementing proper detection of missing configuration variables in
the configuration file.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Add convenience parser/serialization wrappers for all configuration
variables that don't use the default config_parse_*() and
config_serialize_*() helpers yet. This is nothing but a dirty hack and
should be refactored later (e.g. by separating configuration variable
parsing and validation or by adding optional validation functions). It
makes it easier to switch to a more generic configuration parser,
though.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Switch to using the new config_file_walk() helper in config_save() and
update existing configuration variables instead of blindly overwriting
the existing configuration file.
Note: This breaks configuration setting storage if one or more settings
are missing in the configuration file. We need to refactor our parser
and serialization routines another time before this can be implemented
properly, though.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Add config_serialize_conf() which can be used to serialize the value of
a configuration setting (counterpart to config_parse_conf()).
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
This can be used if we care about junk, such as empty lines and
comments. Currently unused since we skip these when parsing
configuration settings. This makes sense if we want to make slight
modifications to the configuration file without losing formatting and
comments, though.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
This adds one level of abstraction to config_load() by splitting out the
actual reading routine and the variable setter into two separate
functions. config_file_walk() can be used to read the configuration
file, strip comments and pass every key/value pair to a callback.
config_load_cb() is the new callback used in config_load().
Rationale: It makes sense to reuse the key/value parser to allow for a
much saner config_save() routine that changes single values only instead
of rewriting (and overwriting) the whole configuration file every time.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Make sure we don't return bogus list elements if negative indexes are
used in llist_{,find_}nth(). Bail out early and return NULL instead.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
* search-001.sh: Compare the output of "-S <regex>" (using a regular
expression) with hardcoded data.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
* day-001.sh: Compare output of "-d<date>" with hardcoded data.
* day-002.sh: Compare output of "-d<num>" with hardcoded data.
* day-003.sh: Compare output of "-d<num>" with "-s<date> -r<num>".
* range-001.sh: Compare output of "-r" with hardcoded data.
* range-002.sh: Compare output of "-r<num>" with hardcoded data.
* range-003.sh: Compare output of "-r<num>" with "-s<date> -r<num>".
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
* todo-001.sh: Tests the "-t" command line option (calculates expected
output by parsing the todo data file).
* todo-002.sh: Tests "-t<num>" in a way similar to todo-001.sh.
* todo-003.sh: Tests "-t0" in a way similar to todo-001.sh.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
* true-001.sh: Always returns true - this should never ever fail.
* run-test-001.sh: Invokes run-test, passing itself as a parameter.
Subsequently, it outputs the very same (hardcoded) expected and actual
values.
* run-test-002.sh: Tests run-test's negative assertion feature in a way
similar to run-test-001.sh. However, output different (hardcoded)
expected and actual values and invoke run-test with the negative test
prefix ('!').
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Sometimes, we might want to make negative assertions (tests where
expected and actual output are expected/known to be different). A test
can be marked negative by prefixing it with an exclamation mark ('!'):
$ ./run-test !test-negative
Running test-negative... ok
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Introduce a new "test/" sub-directory that contains tests for calcurse.
Right now, it only includes the quick-and-dirty "run-test" helper that
can be used to run and verify tests:
$ ./run-test test-1 test-2 test-3 test-4
Running test-1... ok
Running test-2... ok
Running test-3... FAIL
Each argument passed to run-test must be a test script located in the
current directory. run-test invokes each script twice and passes the
command line argument "expected" and "actual", respectively. A test case
succeeds if both "expected" and "actual" instances return with a zero
exit status and produce exactly the same output. It fails otherwise.
run-test terminates with a non-zero exit status as soon as one of the
test fails.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We added count prefix support to motion commands in commit
59e006e56d9f893506af56a4ca114fe53b537e49 but obviously forgot to check
whether the range checks in calendar_move() still work correctly.
Refactor out range checks and replace them by a single check that is
performed *after* the new date is computed, but before we assign the new
value to the actual selected date. This ensures we won't have to bother
about these again when changing/adding functionality in/to
calendar_move(), while keeping performance (date_change() is pretty
cheap and range violations are corner cases anyway).
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We currently use this in one place only but might reuse this a couple of
times later (when migrating to libical).
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Extract iCal and pcal import/export routines into separate files. This
reduces complexity of the super huge "io.c" source file and makes it
easier to follow changes that affect the iCal and pcal routines only
(commits affecting both formats are very uncommon).
Before:
$ wc -l src/io.c
2938 src/io.c
After:
$ wc -l src/{io,ical,pcal}.c
1445 src/io.c
1263 src/ical.c
317 src/pcal.c
3025 total
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Commit 4ff3bb9d introduced a new format for parsing durations, thus
allowing a larger string to be retrieved from user input.
However, the string used is still declared with the old length,
leading to crashes when compiling with stack-smashing protection
features (the default on Archlinux). Inputting a duration string of
more than 8 characters (such as "+1d11h11m") would crash calcurse with
a *** stack smashing detected *** message.
Using a larger string from the start fixes the bug.
Signed-off-by: Baptiste Jonglez <baptiste@jonglez.org>
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
Use days/hours/minutes/seconds instead of seconds in the DURATION field.
It fixes interaction with other softwares like phpicalendar.
Lukas: Remove use of dur-week. RFC 5545 states that you cannot use this
in conjunction with dur-day. Also, fix formatting and use proper
constants.
Signed-off-by: Jerome Pinot <ngc891@gmail.com>
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We used different naming schemes in versions prior to 1.8. Given that
calcurse 1.8 has been released more than 4.5 years ago, remove the
legacy code that still handles these. Users upgrading from <1.8 to 3.0.0
might need to convert the appropriate config file variable manually.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
This patch kind of reverts what we did in commit
6377582841118688aee13aff98c9216403582e45. We were a tad off-base there
since using a map doesn't improve maintainability, really. Using
strcmp() at a central location seems perfectly fine and doesn't have the
overhead of the map scanning algorithm.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
We used custom_load_conf() to load the configuration file and
io_save_conf() to save configuration. Move these functions, including
all helpers, to a central location.
Signed-off-by: Lukas Fleischer <calcurse@cryptocrack.de>
