Returning a single struct which contains multiple named fields reduces
the chance of a mistake in the calling code, as compared to returning
multiple unnamed values which are all of the same type.
This enables the config to be reused in multiple binaries - eg. the
upcoming vsp admin binary which will be responsible for writing a
default config file for new vspd deployments.
Upcoming changes constitute breaking public API changes to both the
client and types modules, therefore this bumps the version numbers of
both modules and adds local replacements to go.mod files such that the
new versions can be used before they are publicly tagged.
Run tests and linters against a hard-coded set of local submodules
instead of attempting to find submodules dynamically.
While this has the obvious drawback of needing to manually update the
list of submodules, it greatly simplifies the script by removing a bunch
of regexes and string manipulation. This trade-off seems worthwhile
because the list of submodules in this repo will not be something which
changes often.
This change makes the script less brittle because it is hard-coded to
always run against the local code, regardless of any changes to module
versionining or project dependencies.
Loading config and preparing loggers are two separate concepts which
should be handled individually. Nesting one inside the other makes
little sense and complicates reusing the code.
There is no need to declare a dozen different "default..." vars only to
immediately insert them into a default config struct when the struct
itself could just be initialized with the default values instead.
This removes the --configfile option from vspd. It introduced quite a
few weird edge cases (and a fair bit of code to deal with them) but
afaik nobody actually used it. Note that the --homedir option stays, so
it is still possible to run vspd with config in a non-default location
if required.
Checking for --help as an explicit step before parsing any other configs
makes the code more intuitive by removing a convoluted bit of error
handling, which happened to be unnecessarily duplicated in three places.
Moving it to a function in the internal package makes it reusable by
multiple binaries.
This also enables the IgnoreUnknown option to be used whilst parsing for
help, which ensures the presence of --help will always result in the
help message being printed. This fixes a minor inconsistency where the
help message would be printed if the flag was placed before an invalid
config, but placing it after would cause an invalid config error to be
written instead. For example, `vspd --help --fakeflag` vs `vspd
--fakeflag --help`.
This downgrade changes StakePoolTicketFee back to the version which does
not consider DCP-0012 activation.
This resolves an issue where Decrediton sometimes fails to pay VSP fees,
caused by Decrediton and vspd independently calculating the fee amount
using different versions of the algorithm.
Releasing the new algorithm will need to be more carefully coordinated,
potentially requiring both client and server sides to be updated in
sync.
If broadcasting parent transaction of a ticket fails because it
references unknown outputs, there is a good chance that waiting a few
seconds will resolve the issue because the ancestor transactions will
propagate through the network and reach the mempool of the local dcrd
instance.
A new tab on the admin page displays a list of all tickets which were
registered with the VSP but missed their votes. Clicking on the ticket
hash redirects to the Ticket Search tab with the details of the missed
ticket displayed.
When invoked with -v, curl actually outputs a message to indicate that
this extra param is unnecessary:
"Note: Unnecessary use of -X or --request, GET is already inferred."
The more verbose name was helpful in the past when some code was
handling more than one waitgroup, however that is no longer the case due
to refactoring so reverting to the more concise name makes sense.
Create the shutdown context earlier so process begins handling shutdown
signals earlier. Specifically, the context should be created before
creating any resources which need a graceful shutdown such as the
database. This is important because without explicit signal handling the
process will terminate immediately upon receiving a signal without
running deferred tasks.
