No pb by me, separating them right now. On Sat, Feb 14, 2026 at 10:48 AM Willy Tarreau <[email protected]> wrote:
> Hi David, > > On Thu, Feb 12, 2026 at 05:20:08PM +0000, David Carlier wrote: > > From ef77967e671d523c03ef9c5a5dd2e87ca35dab37 Mon Sep 17 00:00:00 2001 > > From: David Carlier <[email protected]> > > Date: Thu, 12 Feb 2026 17:11:01 +0000 > > Subject: [PATCH] BUG/MEDIUM: deviceatlas: fix resource leaks, off-by-one > and > > config parser issues > > > > Fix several issues in the DeviceAtlas addon: > > > > - Config parsers da_log_level() and da_cache_size() missing return on > > error, causing fall-through on invalid values. > > - Resource leaks on init and hot-reload error paths (atlasimgptr, cnew, > > da_fini). Add NULL checks on strdup() results. > > - Off-by-one in da_haproxy_conv() truncating last UA character. > > - Cookie vlen using wrong length after extraction. > > - Double-checked locking race in hot-reload checkinst path. > > - da_fini() called unconditionally in deinit even when never initialized. > > Removed erroneous shm_unlink() that could affect dadwsch. > > - Increase DA_MAX_HEADERS to 32 and hbuf to 64 for current data files. > > - Precompute maxhdrlen to skip oversized headers early in fetch path. > > - Set cache_size on hot-reloaded atlas instance. > > > > This should be backported to lower branches. > > Why is it that so many independent changes are merged into the same > patch ? Do you still have them separate or can you please split them? > It's not a problem to have 10 patches if required, but at least each > one should do one thing and explain why. It would also help with > backports and make sure that a culprit could be spotted in case of > any possible regressions. > > Thanks! > Willy >
