There are circumstances where input via --appDataDir
will lead to the hiddenservice-directory to be deleted.
Successfully reproduced using
--appDataDir=~/foo
Although the "~" does not get interpreted correctly on
my linux system, it does manage to throw off the
mechanics of sparing the hiddenservice-directory from
being deletd.
Add toProtoMessageSynchronized() default method to PersistableEnvelope,
which performs (blocking) protobuf serialisation in the user thread,
regardless of the calling thread. This should prevent data races like
the ConcurrentModificationException observed in #3752, under the
reasonable assumption that shared persistable objects are only mutated
in the user thread.
Also add a ThreadedPersistableEnvelope sub-interface overriding the
default method above, to let objects which are expensive to serialise
(like DaoStateStore) be selectively serialised in the 'save-file-task-X'
thread as before, but directly synchronised with each mutating op. As
most objects are cheap to serialise, this avoids a noticeable perf drop
without having to track down every mutating method for each store.
In all cases but one, classes implementing ThreadedPersistableEnvelope
are stores like TradeStatistic2Store, with a single ConcurrentHashMap
field. These require no further serialisation, since the map entries are
immutable, so the only mutating operations are map.put(..) calls which
are already synchronised with map reads. (Even if map.values().stream()
sees updates @ different keys happen out-of-order, it should be benign.)
The remaining case is DaoStateStore, which is only ever reset or
modified via a single persist(..) call with a cloned DaoState instance
and hash chain from DaoStateSnapshotService, so there is no aliasing
risk from the various DAO state mutations done in DaoStateService and
elsewhere.
This should fix#3752.
Minor change for consistency: narrow the signature of some remaining
such methods, which have return type 'PersistableEnvelope'.
(This excludes some other cases with return type 'NetworkEnvelope'.)
Bisq frequently (once per minute) queries our price nodes for up-to-date
fee information. It does so by HTTP GET request. However, it provided
a UID via the "User-Agent" HTTP header field. This UID has been a random
number which changed everytime Bisq gets started up.
This UID has never been used. Thus, remove it.
This was originally added with the intention that the local Bitcoin node
port could be customized, but in fact it never could be, because Guice
configuration always hard-wired the value to the default port for the
CurrentBaseNetwork's Parameters (eg. 8333 for BTC_MAINNET).
This change removes the constant, removes any Guice wiring and injection
and localizes the hard-coded assignment to the LocalBitcoinNode
constructor to simplify and make things explicit.
If it is desired to allow users to specify a custom port for their local
Bitcoin node, a proper option shoud be added to Config. In the meantime,
users may work around this by using `--btcNodes=localhost:4242` where
4242 is the custom port. Note however, that the pruning and bloom filter
checks will not occur in this case as the provided node address will not
being treated as a LocalBitcoinNode.
Introduces LocalBitcoinNode::willUse and ::willIgnore to move logic that
was previously littered throughout the codebase into one place. Also,
changes the usages of LocalBitcoinNode to be more precise, specifying
which of these questions are being asked:
- "is there a local BTC node" (isDetected);
- "is it well configured" (isWellConfigured and isUsable);
- "will we ignore a local BTC node even if we found a usable one"
(willIgnore);
- "is there a usable local BTC node and will we use it" (willUse).
These changes make related logic much easier to maintain and to read.
It's quite amazing how obvious this was, yet I missed it for such a long
time. Simplifies usage of LocalBitcoinNode and its internals even more
so. Fixes#4005. The way we structured LocalBitcoinNode was as if the
detection checks were expensive, but they're not. Previously, in some
cases we would notice that a local BTC node wouldn't be used even if it
was detected, so we would skip these checks. This optimization now
doesn't happen. It might be reimplemented in a coming change where more
local BTC node logic is moved into LocalBitcoinNode, but, even if it's
not, this check is fairly cheap. A notable exception is if the local BTC
node is not responding, which would cause us to wait for a timeout, but
if that is the case the mentioned optimization wouldn't help (most of
the time).
As mentioned in the issue #3984, Have created lot of buy/sell offers
with different currencies spanning across around 30+ offers
so that we need to use scroll bar in my open offers and able to
recreate the issue.
This change fixes#3998 by avoiding the attempt to create a user's
appDataDir if it already exists. Normally this attempt is not a problem,
as Files#createDirectories does nothing if the target is a directory
that already exists, but in the case that the target is a symbolic link,
Files#createDirectories throws a FileAlreadyExistsException. This change
prevents the call to Files#createDirectories if the target already
exists, and if in fact the target is a symbolic link pointing to an
existing directory, this is not a problem and everything proceeds as per
usual. This mimics the behavior in Bisq v1.2.5 and prior, where
File#mkdirs was used instead of Files#createDirectories. File#mkdirs
does not throw an exception when the target directory is a symlink, it
just returns false.
This is a pure refactoring that renames and inlines variables to tighten
up and make more consistent the implementation of these two methods. It
is done in preparation for a subsequent substantive change that fixes a
bug.
Previously, Config#mkdir and #mkAppDir threw ConfigException if any
underlying IOException was thrown from Files#createDirectory or
Files#createDirectories respectively. This resulted in only a simple
error message being reported at the command line and the details of the
underlying exception's stack trace were lost. This change wraps and
rethrows the underlying IOException as an UncheckedIOException such that
it gets caught by the catch(Throwable) clause in BisqExecutable, which
does print a stacktrace to stderr. This is the way it always should have
worked; throwing ConfigException in these cases was an oversight in the
original implementation. This change also removes the ConfigException
constructor that allows for passing a nested exception to be wrapped as
there is actually no use case for this now that these two anomalies have
been corrected.
This change was made in the process of attempting to reproduce #3998.
