[ovs-dev] [PATCH] ovsdb-idl: Mark row "parsed" in ovsdb_idl_txn_write__
damjan.skvarc at gmail.com
Fri Jul 19 02:50:39 UTC 2019
On Thu, 18 Jul 2019, 20:12 Ben Pfaff, <blp at ovn.org> wrote:
> I guess I missed it. Will you please point it out to me, for example in
> the list archive?
> On Thu, Jul 18, 2019 at 06:58:34PM +0200, Damijan Skvarc wrote:
> > Hmm, the problem of "parsed" flag is that it identifies "all" columns of
> > certain row have been parsed, however there are CLI tools which modify
> > individual colums by calling ovsdb_idl_txn_write_() function. In this
> > and in case parsed flag would be set in ovsdb_idl_txn_write() function
> > unparsing procedure would be called also for columns which were not
> > The problem could be overcome by having individual flag for each column.
> > This has been addresed in pending pull request. Apparent mail has been
> > to dev list, but obviosly has been somehow overlooked.
> > br, damijan
> > On Thu, 18 Jul 2019, 17:52 Ben Pfaff, <blp at ovn.org> wrote:
> > > On Wed, Jul 17, 2019 at 09:05:04PM +0200, Dumitru Ceara wrote:
> > > > Once a column is set in a row using ovsdb_idl_txn_write__ we also
> > > > the row "parsed". Otherwise the memory allocated by
> > > > sbrec_<table>_parse_<col>() will never be freed. After marking the
> > > > "parsed", the ovsdb_idl_txn_disassemble function will properly free
> > > > newly allocated memory for the column (ovsdb_idl_row_unparse).
> > >
> > > Wow, good catch. I bet that's been there forever.
> > >
> > > The OVSDB IDL code is too complicated. It's too hard to spot the
> > > issues. I wish I saw a way to make it simpler.
> > > _______________________________________________
> > > dev mailing list
> > > dev at openvswitch.org
> > > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
> > >
More information about the dev