On Sun, Nov 01, 2026 at 02:52:22PM +0100, Jeremie Courreges-Anglas wrote: > On Sat, Oct 31, 2026 at 09:19:01AM -0400, Kurt Mosiejczuk wrote: > > On Sat, Oct 31, 2026 at 12:15:30PM +0100, Jeremie Courreges-Anglas wrote: > > > On Fri, Jul 03, 2026 at 05:25:43PM +0100, Stuart Henderson wrote: > > > > On 2026/07/03 18:11, Jeremie Courreges-Anglas wrote: > > > [...] > > > > > Pending some actual migration reports, I'd like to postpone the > > > > > decision of adding @ask-update. Users could probably also benefit > > > > > from a current.html entry, which could very well be enough given the > > > > > availability of mdb_dump-0.9 in the package and the loud MESSAGE. > > > > > > makes sense. > > > > > So far I only got a report from kmos who reported no visible issue, > > > even though I doubt that lwjgl3 is a heavy lmdb user. Therefore I'm > > > heading for just a current.html entry + MESSAGE. > > > > > Here's an updated diff with the following changes: > > > - update to 1.0.2 (and 0.9.36) > > > - move to github since upstream's gitlab refuses me with http 429 > > > - keep PORTROACH from sthen since portroach apparently doesn't handle > > > DIST_TUPLE > > > - drop pointless Makefile hunks, we're overriding the variables on the > > > command-line anyway > > > - use the .pc file generated upstream, but tweak its generation > > > > > I'll run a partial bulk but I strongly suspect nothing will break. > > > Upgrade test reports still welcome. > > > > Well, I would definitel like this to go in because now latest lwjgl3 > > requires 1.0.x. > > > > ok kmos > > Thx for the feedback & ok, Kurt. Here's a tweaked version that > appears needed at least for net/knot. The new SysV semaphores code > path breaks net/knot(-3.6.2, not sure about -3.5.8) with semop(2) > returning EACCES. I don't feel like debugging the root cause so let's > just use the same semaphore implem as previously. > > Some tweaks: > - add patch comments > - use mdb_dump -a in test case and MESSAGE
Using mdb_dump-0.9 -a in the example may be misleading since it errors out when the source database doesn't contain sub-databases ; I'll find another way to mention it in MESSAGE before committing > Except for a bogus code pattern in bogofilter fixed yesterday I'm not > aware of any other regression. I will therefore commit this very soon > unless I hear objections. More test reports & oks are still welcome. Tentative diff for current.html. Index: current.html =================================================================== RCS file: /cvs/www/faq/current.html,v diff -u -p -r1.1166 current.html --- current.html 16 Oct 2026 06:16:59 -0000 1.1166 +++ current.html 2 Nov 2026 15:04:56 -0000 @@ -85,6 +85,19 @@ interface, as its sessions may also be t See <a href="https://man.openbsd.org/pppoe";>pppoe(4)</a> for details. +<h3 id="r202611XX">2026/11/XX - [packages] LMDB 1.0 incompatible database format</h3> + +<p> +LMDB was updated to 1.0. +This version doesn't understand the database format used by LMDB 0.9. +You can use <code>mdb_dump-0.9 [-a]</code> to create a dump of your +existing databases and <code>mdb_load</code> to restore them. +<pre class="cmdbox"> +mdb_dump-0.9 old-db > old-db.txt +mdb_load -f old-db.txt new-db +</pre> + + <!-- New sections at the bottom, with two blank lines before. New sentences start on new lines. -- jca
