Builds made with ghc 9.10+ and the 'debug' build flag, will show
(some kind of, partial) stack traces when the program ends with an
error. (And also will have ghc-debug support enabled.)
The stack traces will probably improve in due course.
A followup to #2099, #2100 and #2127. Now relative links to js/css
resources will use the same hostname etc. the main page was requested
from, making them work better when accessed via multiple IP
addresses/hostnames without an explicit --base-url setting.
It's clearer to write command-specific flags after the command name
argument, but that's no longer required.
(Writing non-builtin, addon-specific flags after -- is still required).
Also, give up on "obey help/doc/version flags even if there's a bad
command/flag", it's too hard to do well.
When built with the ghcdebug flag and started with --debug=-1 (or -2
to pause at startup, or -3 to pause before exit), hledger can be
controlled by ghc-debug clients like ghc-debug-brick or a custom
ghc-debug query script.
Also, refactor version string code.
Changes:
1. rename the sandstorm "manage" permission to "edit"
(old permission names: view, add, manage;
new permission names: view, add, edit).
Rationale: "edit" best describes this permission's current powers, to users and to operators.
If we ever added more manager-type features we'd want that to be a new permission,
not a rename of the existing one (which would change the powers of existing users).
2. rename the sandstorm roles for consistency with permissions
(old role names: viewer, editor, manager;
new role names: viewer, adder, editor)
Rationale: it's needed to avoid confusion.
3. add a new option: --allow=view|add|edit|sandstorm (default: add).
'sandstorm' sets permissions according to the X-Sandstorm-Permissions header.
Drop the --capabilities and --capabilities-header options.
Rationale: it's simpler and more intuitive.
4. replace "capability" with "permission" in ui/docs/code.
Rationale: consistent with the above, more familiar.
We no longer rely on git tags and git describe output, since it's
hard to reliably select the correct tag eg for minor releases.
We might use them again in future for dev builds, but it requires
adding git describe --match support to githash. For now,
* Program name, OS and architecture are always shown.
* The package version is always shown.
* If there is git info at build time, the latest commit hash and commit date are shown.
Example outputs:
A homebrew binary, not built in git repo: hledger-ui 1.24, mac-aarch64
A CI release build, built in git repo: hledger 1.24.1-g455b35293-20211210, mac-x86_64
API changes:
* new type synonyms ProgramName, PackageVersion, VersionString
* versionStringForProgname -> versionString with extra argument
* versionStringFor -> versionStringWith with extra argument
to Hledger.Cli.
Since the git hash changes whenever any file in the repository changes,
this means Hledger.Cli.Version needs to be recompiled all the time.
Since it is at the bottom of the module hierarchy, this means that the
whole hledger package needs to be recompiled. We instead move the
TemplateHaskell splice to one of the top modules, so much less needs to
be recompiled.
Note: Ghc seems to be able to get out of most of the recompiling a lot
of the time (due to caching?), but this makes things more reliable.
This commit introduces IPv6 support (and thus closes#1145).
It also allows using local hostnames as a parameter for --host.
For this, multiple things needed to be changed:
- checkWebOpts is dropped, as the supplied parameter is checked later
by Network.Socket.getAddrInfo
- defbaseurl needs to check if : is used in the host, as this indicates
the usage of an IPv6 address. In this case, the host needs to be
wrapped in [] for the base URL
- To allow opening such a modified base URL, runHostPortFullUrl is used
instead of runhostPortUrl, as it allows opening arbitrary URLs instead
of a path prefixed with http://127.0.0.1
As checking the host for validity is postponed until the webserver tries
to start, an invalid hostname leads to an exception caused by
Network.Socket.getAddrInfo.
This is still fine, as hledger-web won't start in an undefined state, but
will terminate with a nonzero exit code.
This commit adds the --socket option to use hledger-web over an AF_UNIX socket
file.
It allows running multiple instances of hledger-web on the same system without
having to manually choose a port for each instance, which is helpful for running
individual instances for multiple users. In this scenario, the socket path is
predictable, as it can be derived from the username.
It also introduces the following dependencies:
- network
- Used to create the unix domain socket
- unix-compat
- Used to identify if the socket file is still a socket, to reduce the risk
of deleting a file when cleaning up the socket
- Modified the cors option to require a String
- Moved the logic to build the cors policy to WebOptions.hs
- Specify the --cors "*" example in the cors option help
- Added utf8-string dependency to convert a String into a ByteString
These Ledger-ish spellings were dropped over the years, to improve
--help's layout. Now they work again, but are considered hidden flags
so --help doesn't show them automatically (but I decided to mention
them manually in --date2's description).
