If a connection does e.g. CREATE TABLE, it will start a "child statement"
to reparse the schema. That statement does not start its own transaction,
and so should not try to end the existing one either.
We had a logic bug where these steps would happen:
- `CREATE TABLE` executed successfully
- pread fault happens inside `ParseSchema` child stmt
- `handle_program_error()` is called
- `pager.end_tx()` returns immediately because `is_nested_stmt` is true
and we correctly no-op it.
- however, crucially: `handle_program_error()` then sets tx state to None
- parent statement now catches error from nested stmt and calls
`handle_program_error()`, which calls `pager.end_tx()` again, and since
txn state is None, when it calls `rollback()` we panic on the assertion
`"dirty pages should be empty for read txn"`
Solution:
Do not do _any_ error processing in `handle_program_error()` inside a nested
stmt. This means that the parent write txn is still active when it processes
the error from the child and we avoid this panic.
Previously, the encryption module had hardcoded a lot of things. This
refactor makes it slightly nice and makes it configurable.
Right now cipher algorithm is assumed and hardcoded, I will make that
configurable in the upcoming PR
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2722
This PR make it possible to do 2 pretty crazy things with turso-db:
1. Now we can mix WAL frames inserts with SQL execution within same
transaction. This will allow sync engine to execute rebase of local
changes within atomically over main database file (the operation first
require us to push new frames to physically revert local changes and
then we need to replay local logical changes on top of the modified DB
state)
2. Under `conn_raw_api` Cargo feature turso-db now expose method which
allow caller to specify WAL file path. This dangerous capability exposed
for sync-engine which maintain 2 databases: main one and "revert"-DB
which shares same DB file but has it's own separate WAL. As sync-engine
has full control over checkpoint - it can guarantee that DB file will be
consistent with both main and "revert" DB WALs.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2716
Add `truncate` method in the page cache which remove all entries which
reference pages greater than new DB size.
This will be used in the sync engine as in its case DB size can shrink
when we "rebase" changes from remote to local.
It stands on the #2707 because touch few files from that PR
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2711
The Unix backend is a syscall()-based, blocking implementation. The
O_NONBLOCK adds nothing.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2708
This PR adds information about checkpoint sequence number to the WAL raw
API. Will be used in the sync engine.
Depends on the #2699
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2707
SQLite does not store the rowid alias column in the record at all
when it is a rowid alias, because the rowid is always stored anyway
in the record header.
Found when running simulator in #2641
All indexes store the rowid as the last column, so whenever the rowid of
a given row changes the index entry must also be deleted and reinserted
with the new index.
Reviewed-by: Nikita Sivukhin (@sivukhin)
Closes#2712
UPDATE should skip over the UNIQUE constraint failure if the existing
row it found during the check has the same rowid as the row we are
currently updating
Same deal as #2700, except this time in UPDATE. Nothing tests this on
`main` so not caught.
I will later put #2641 into mergeable condition so it will catch all of
these going forward.
Reviewed-by: Nikita Sivukhin (@sivukhin)
Closes#2710
Previously, we just hardcoded the reserved space with encryption flag.
This patch removes that and sets the reserved space if a key was
specified during a creation of db
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2706
