nmcli has documentation strings embedded. Those strings are extracted
from gtk-doc comments, using pygobject and put in the generated file
"clients/common/settings-docs.c".
This file "clients/common/settings-docs.c" is disted, so from
a source tarball you can build nmcli without enabling introspection.
However, when building from a git-tree, the file is missing and
thus one cannot build --with-nmcli unless also using at least
--enable-introspection to generate "clients/common/settings-docs.c".
That is inconvenient. Especially during cross-compilation, where
one also needs python and pygobject in the foreign architecture (because
the generation of "settings-docs.c" loads the built libnm.so via
pygobject). It is bad because nmcli is an essential part of
NetworkManager, so building --without-nmcli is not a great option.
Previously, the only alternative was to pre-generate a source tarball
on a separate machine and build that. This however complicates efforts
to automatically build git snapshots of NetworkManager.
Fix that by commiting "clients/common/settings-docs.c.in" to git.
When building with --disable-introspection, the pre-generated
file is used instead. This is fine, because the file only depends
on static, checked-in documentation strings that seldomly change.
Also add a check target to notice when the pre-generated file differs
from what we are about to generate during --enable-introspection.
That happens when editing one of the gtk-doc entires. In this case,
`make check` will notify that the pre-generated "settings-docs.c.in"
file needs updating too.
Yes, when changing gtk-doc comments you need to updte the file manually.
At least, the check failure notifies you.
Depending on the get_type argument, we don't only want
to return strings, but arbitrary pointers.
The out_to_free argument still makes sense, but depending on
the get-type you must know how to free the pointer.
Currently we only have two get-types: PRETTY and PARSABLE.
In the future we may want to add more of those, so the
default behavior when encountering an unrecognized get-type
should be PARSABLE.
Don't ever check whether get-type is PARSABLE. Check instead,
whether it is PRETTY (the non-default) or do the default (PARSABLE).
When generating output data, nmcli iterates over a list of
property-descriptors (nmc_fields_ip4_config), creates an intermediate
array (output_data) and finally prints it.
However, previously both the meta data (nmc_fields_ip4_config) and
the intermediate format use the same type NmcOutputField. This means,
certain fields are relevant to describe a property, and other fields
are output/formatting fields.
Split this up. Now, the meta data is tracked in form of an NMMetaAbstractInfo
lists. This separates the information about properties from intermediate steps
during creation of the output.
Note that currently functions like print_ip4_config() still have all the
knowledge about how to generate the output. That is wrong, instead, the
meta data (NMMetaAbstractInfo) should describe how to create the output
and then all those functions could be replaced. This means, later we want
to add more knowledge to the NMMetaAbstractInfo, so it is important to
keep them separate from NmcOutputField.
Embed a @meta_type structure in NMMetaSettingInfoEditor and
NMMetaPropertyInfo. This allows to make the NMMeta*Info instances
themself to become generic and they can be passed around as generic
NMMetaAbstractInfo types.
For one, the embedded NMMetaType pointer can be used to determine
of which type a NMMetaAbstractInfo instance is. On the other hand,
the NMMetaType struct can be extended to be a VTable to provide
generic access to the type.
In the end, both NMMetaSettingInfoEditor and NMMetaPropertyInfo are
conceptionally very similar: the describe a certain type and provide
accessors.
In nmcli we have yet another NMMetaAbstractInfo type: NmcOutputField
will be modified to become another implementation of meta data (it
already is, it just cannot be used interchangable with the other
types).
Also, embed the NMMetaSettingInfoEditor in the NMMetaPropertyInfo
instance. This allows from a given NMMetaPropertyInfo to retrieve it's
parent NMMetaSettingInfoEditor.
"nm-meta-setting-desc.h" contains static type description, vtable and (internal)
accessor functions. Add accessor functions that operate on top of the type description
to "nm-meta-setting-access.h".
This check requires additional information about the environment, that
is about the present connections in NMClient.
"nm-meta-setting-desc.c" should be independent from the libnm D-Bus
cache, hence move this code to "settings.c".
The lower layers are concerned with handling settings. They should not
be aware of how to notify about warnings. Instead, signal them via
the warn_fcn() hook.
For G_TYPE_BOOLEAN, let it get handled by the getter hook instead
of modifying system-wide behavior of glib.
Also, there are no properties of G_TYPE_CHAR. Just drop that.
Options dependant on specific commands (e.g., nmcli connection show
--active) are now allowed to be processed by the next_arg() function.
This would allow autocompletion to expand options belonging to specific
command first, and then global ones.
Note that global options ("--ask" and "--show-secrets") will be auto-completed
everywhere but only if at least a '-' is passed. Command specific ones
(--temporary, --active, --order) will be auto-completed only after the command
they belongs to but without requiring the user to pass a heading '-'.
Example:
'nmcli connection show -a'
will expand '-a' into '--active', but
'nmcli connection add -a`
will expand '-a' into '--ask' (as it is a global option)
This commit fixes also autocompletion for:
nmcli connection modify --temporary
This happens when the connection is in "activating" state and
the connection is a master one waiting for slaves: "check_activated()"
is called by the active_connection_state_cb() and device_state_cb() callbacks.
If the device has already moved to a state >= NM_DEVICE_STATE_IP_CONFIG, the
call to active_connection_state_cb() will end calling
activate_connection_info_finish(), freeing the "info" object. The subsequent
call to device_state_cb() will result in accessing the freed "info".
Just call check_activated() once after registering the
active_connection_state() and device_state_cb() callbacks.
Before commit a63c4d0824 ("cli: use designated initializers for
setting NmcOutputField fields") each field had a @name and a
@name_l10n , which were equal for all properties except for wifi
IN-USE:
{"IN-USE", N_("*")}, /* 15 */
The commit removed @name_l10n so now the displayed name is equal to
the field name. But now:
$ nmcli device wifi list
Error: 'device wifi': invalid field 'IN-USE'; allowed fields:
NAME,SSID,SSID-HEX,BSSID,MODE,CHAN,FREQ,RATE,SIGNAL,BARS,SECURITY,
WPA-FLAGS,RSN-FLAGS,DEVICE,ACTIVE,*,DBUS-PATH
The new field name should be 'IN-USE' and not '*', otherwise scripts
doing "-f IN-USE" will break. As a side effect we now show 'IN-USE'
instead of '*' in the column header, which seems easier to understand:
IN-USE SSID MODE CHAN RATE SIGNAL BARS SECURITY
* default Infra 5 54 Mbit/s 71 ▂▄▆_ WPA2
guest Infra 1 54 Mbit/s 62 ▂▄▆_ WPA2
Fixes: a63c4d0824
We should not violate the global data to track the output data
while it is constructed and printed.
Most of the time, we actually clear the output data anyway --
either before constructing it, or after printing it.
In some cases we didn't, but I think that is a bug. It's really
hard to keep track of this.
The output data should belong to a certain scope and get destroyed
afterwards. Passing it around is very confusing. Don't do that.
To better understand which part of the code have side effects,
split print_data() in a part that mutilates the input array
and a part that only prints.
Don't cast const strings to non-const. And don't track
whether to free a variable in a boolean. Instead, assign
ownership to variables that get destroyed when the function
returns.
The NmCli structure is passed around everywhere and contains all
the state of the program. It is very hard to follow which parts
are used where.
Split out more configuration options to a NmcConfig field. This field
is mostly immutable, and modified at particular places that now can be
easily found.
nmc contains all options and collects output data. It is very hard to
understand what it does. Start splitting it up, and pass arguments along
-- as needed.
Otherwise, changing the structure is difficult because it all depends
on the order (and you don't immdiately see which field is used where).
Also, drop the name_l10n field.
These functions are only used by nm-meta-setting-desc.c. Make them internal.
Unfortunately, they are part of "common.h" which cannot be used without
the rest of nmcli. Still todo.
This part contains static functions and variables to describe
settings. It is distinct from the mechanism to use them, or
access them.
Split it out.
It still uses clients/cli/common.h and clients/cli/utils.h
which shall be fixed next.
They will be moved out of nmcli as they are generally useful.
Even if they happen to be used only inside nmcli, there should
be a better separation between logic (nmcli) and setting decriptions.
Mostly these strings are static. In same cases they are generated
however. Instead of caching them in a static variable, return them
to the caller.
And belatedly fix invoking describe_fcn().
