Bugzilla

The revised snapshot is up.  I pushed the revised snapshot and a simple patch for TelemetryVFS to try:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=493d9d030cbad5e4a0ba43e5d157fce381c63c13

Linux distributions that build with system SQLite potentially mean that we'll need to make a more complicated version of the patch that performs runtime detction of the SQLite version.  (Unless those distributions make sure to rebuild Firefox every time they rebuild SQLite?)

This might look like leaving the method as it exists intact and having it check the version and early return zName if sqlite3_uri_parameter is known to have the new behavior.  This patch would potentially want to be uplifted so that Firefox doesn't break when the new SQLite version gets rolled out by distributions.

Then when we can make it a hard build-time dependency to require 3.31.0, we can remove the method and bake the requirement into configure.in.  Or maybe we can do that already on trunk and we should just uplift the runtime approach to beta?

The revised snapshot is up.  I pushed the revised snapshot and a simple patch for TelemetryVFS to try:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=493d9d030cbad5e4a0ba43e5d157fce381c63c13

Linux distributions that build with system SQLite potentially mean that we'll need to make a more complicated version of the patch that performs runtime detction of the SQLite version.  (Unless those distributions make sure to rebuild Firefox every time they rebuild SQLite and have a hard version dependency, in which case we can use ifdef's.)

This might look like leaving the method as it exists intact and having it check the version and early return zName if sqlite3_uri_parameter is known to have the new behavior.  This patch would potentially want to be uplifted so that Firefox doesn't break when the new SQLite version gets rolled out by distributions.

Then when we can make it a hard build-time dependency to require 3.31.0, we can remove the method and bake the requirement into configure.in.  Or maybe we can do that already on trunk and we should just uplift the runtime approach to beta?