[sqlite] Broken database after experiments with fts.

2016-04-09 Thread Cezary H. Noweta
Hello, On 2016-03-29 19:25, Dan Kennedy wrote: >http://sqlite.org/src/info/d0a3853b37230c12 > The fix should appear as part of the 3.13 release (not 3.12 - the > version that was tagged earlier this morning). Could that check--in be incorporated into next minor version? It fixes rather seri

[sqlite] Broken database after experiments with fts.

2016-03-30 Thread Dan Kennedy
On 03/30/2016 12:14 AM, Cezary H. Noweta wrote: > Hello, > > On 2016-03-26 15:37, John Found wrote: >> Why cannot drop the table test? >> >> sqlite> begin transaction; >> sqlite> create virtual table test using fts5; >> Error: vtable constructor failed: test >> sqlite> commit; > > It is not require

[sqlite] Broken database after experiments with fts.

2016-03-29 Thread Cezary H. Noweta
Hello, On 2016-03-29 20:36, Cezary H. Noweta wrote: > I'm sorry, if I've missed something, but what about Parse::isMultiWrite? > AFAIK, Parse::mayAbort is used solely in conjunction with > Parse::isMultiWrite. How can lone Parse::mayAbort influence on > OP_Transaction (or on any other behavior re

[sqlite] Broken database after experiments with fts.

2016-03-29 Thread Cezary H. Noweta
Hello, On 2016-03-29 19:25, Dan Kennedy wrote: > I think that's the right thing to do as well. See here: > >http://sqlite.org/src/info/d0a3853b37230c12 > > The fix should appear as part of the 3.13 release (not 3.12 - the > version that was tagged earlier this morning). I'm sorry, if I've mi

[sqlite] Broken database after experiments with fts.

2016-03-29 Thread Cezary H. Noweta
Hello, On 2016-03-26 15:37, John Found wrote: > Why cannot drop the table test? > > sqlite> begin transaction; > sqlite> create virtual table test using fts5; > Error: vtable constructor failed: test > sqlite> commit; It is not required. Non-commited-and-non-rolledback, dangling transaction supp

[sqlite] Broken database after experiments with fts.

2016-03-26 Thread John Found
On Sat, 26 Mar 2016 16:08:33 +0100 Eduardo Morras wrote: > On Sat, 26 Mar 2016 16:37:18 +0200 > John Found wrote: > > > Why cannot drop the table test? > > > > sqlite> begin transaction; > > sqlite> create virtual table test using fts5; > > Error: vtable constructor failed: test > > sqlite> co

[sqlite] Broken database after experiments with fts.

2016-03-26 Thread John Found
Why cannot drop the table test? sqlite> begin transaction; sqlite> create virtual table test using fts5; Error: vtable constructor failed: test sqlite> commit; sqlite> sqlite> drop table test; Error: vtable constructor failed: test sqlite> .tables test test_content test_docsize test_co

[sqlite] Broken database after experiments with fts.

2016-03-26 Thread Eduardo Morras
On Sat, 26 Mar 2016 16:37:18 +0200 John Found wrote: > Why cannot drop the table test? > > sqlite> begin transaction; > sqlite> create virtual table test using fts5; > Error: vtable constructor failed: test > sqlite> commit; > sqlite> > sqlite> drop table test; > Error: vtable constructor faile

[sqlite] Broken database after experiments with fts.

2016-03-26 Thread Keith Medcalf
And only when commit is used to end the transaction, an explicit rollback works properly. > On Saturday, 26 March, 2016 09:09, Eduardo Morras > wrote: > > Clearly the tables were created, however (see the output from .tables). > > This *only* occurs if you wrap the "create virtual table" in a

[sqlite] Broken database after experiments with fts.

2016-03-26 Thread Keith Medcalf
On Saturday, 26 March, 2016 09:09, Eduardo Morras wrote: Clearly the tables were created, however (see the output from .tables). This *only* occurs if you wrap the "create virtual table" in an explicit transaction but does not occur if an implicit transaction is used. The tables are persist