| Age | Commit message (Collapse) | Author | Files | Lines |
|
Git for windows is used to checkout the repository and it
creates broken symlinks (files containing the target) which
break the tests because e.g. the lexer.lua module can not
be loaded.
We use a bash script to convert these pseudo symlinks:
http://stackoverflow.com/questions/38140373
and invoke it via powershell because that seems to be the
only way to execute multiline commands from within the
AppVeyor YAML configuration file.
|
|
|
|
|
|
|
|
|
|
Also include all *.exe files into the bulid artifact.
|
|
|
|
|
|
Still hopefully that it will eventually become accessible again.
After github reverts the flagging of my account. For now I have cloned
it locally just in case.
|
|
The test/core tests use source files from the repository root
(e.g. ../../text.c) which seems to confuse gcov and/or codecov.io
because the same file is used under multiple paths.
We add a new entry to the test matrix which just runs the core/tests.
This is done by invoking the `make clean` target and skipping any
tests which require the main vis binary.
|
|
This will most likely break macOS travis builds.
|
|
This results in a slightly smaller binary while still avoiding #ifdefs.
Close #494
|
|
These do not really belong into the editor core. If desired they
could be implemented in Lua instead.
|
|
|
|
|
|
This should fix build breakage of the test/util/keys utility.
|
|
Try to test the different regex backends and builds with/without Lua support.
|
|
|
|
|
|
They will still not be displayed correctly, but at least they
should no longer cause memory errors.
|
|
Should fix travis failure.
|
|
|
|
Unlike vim we do not respect `:set expandtab` here.
|
|
Fix #491
|
|
|
|
Strictly speaking this is a compiler bug:
https://llvm.org/bugs/show_bug.cgi?id=22062
The C11 standard section 6.4.4.3 says:
"An identifier declared as an enumeration constant has type int."
and 6.7.2.2:
"Each enumerated type shall be compatible with char, a signed
integer type, or an unsigned integer type. The choice of type is
implementation-defined, but shall be capable of representing
the values of all the members of the enumeration."
So while `err` can store a value larger than that of any enumeration
member, it could also be of signed type, resulting in a warning about
comparing integers of different signs.
Converting it to size_t before the range check and array indexing,
should fix both warnings.
Fix #478
|
|
|
|
|
|
By now we have efficient forward searches using the TRE regex backend.
Further improvements can be discussed in #488.
Close #260
|
|
Fix #487
|
|
|
|
Fixes CID 141179
|
|
Fixes CID 137371
|
|
Fixes CID 141180
|
|
|
|
We now guarantee the existence of at least one non-sentinel
piece at all time and allow iterators to recover from these
delimiting pieces.
|
|
In an ideal world this should never happen. However,
if for some reason all cursors/selections become
invalid, default to an empty range. This at least
allows to terminate the editor gracefully whereas
before the command would silently be ignored.
|
|
Unlike EOF address 0 needs no special treatment.
|
|
This should fix display issues when entering the first newline of a file.
|
|
|
|
This allows lines like
: ${FOO:="${bar}/baz"}
to be highlighted correctly.
|
|
|
|
Fix `gN` and delimited inner variants when given a count.
|
|
|
|
The character following a movement_key command should not be subject to
keymap translation since it is used to find characters in the document.
|
|
|
|
|
|
|
|
Previously this was done implicitly through the primary cursor
which was always adjusted when a new cursor is being created.
However, this is no longer the case and we do not want to iterate
through all cursors when creating a new one.
In the longterm we might want to store cursors in a contiguous
memory location (i.e. an array) instead of chaising pointers
all over the place.
|
|
We currently have the invariant that the primary cursor is always
placed within the visisble viewport.
Previously view_cursors_new would automatically make the new cursor
primary. This in turn causes the viewport to be adjusted triggering
lots of unnecessary redraws. As a result commands creating many new
selections might become unbearably slow.
Instead the caller has to explicitly make the new cursor primary.
|