Open dridi opened 8 months ago
Bugwash:
SLT_DeliveryError
bugwash:
-g session
, it is not particularly usefulSLT_Error
is not a particularly good tag for this, because it is currently used for internal errors which can usually be addressed with VCL or changing parametersFetchError
, should add a DeliveryError
tag and then use the former consistently for VFPs, and the latter for VDPs
SLT_Error
is not a particularly good tag for this, because it is currently used for internal errors which can usually be addressed with VCL or changing parameters
nitpick: this can be addressed with send_timeout
for that case, but point taken, there is nothing we can do when the client disconnects.
I'd like to cover this during bugwash today, before we send this to production.
bionic failure,
cache/cache_expire.c: In function 'exp_inbox':
cache/cache_expire.c:282:61: error: format '%ld' expects argument of type 'long int', but argument 6 has type 'VCL_INT {aka long long int}' [-Werror=format=]
VSLb(&ep->vsl, SLT_ExpKill, "EXP_Removed x=%ju t=%.0f h=%ld",
~~^
%lld
cache/cache_expire.c:284:7:
oc->hits);
~~~~~~~~
cache/cache_expire.c: In function 'exp_expire':
cache/cache_expire.c:360:61: error: format '%ld' expects argument of type 'long int', but argument 6 has type 'VCL_INT {aka long long int}' [-Werror=format=]
VSLb(&ep->vsl, SLT_ExpKill, "EXP_Expired x=%ju t=%.0f h=%ld",
~~^
%lld
cache/cache_expire.c:362:7:
oc->hits);
~~~~~~~~
CC cache/varnishd-cache_fetch_proc.o
cache/cache_expire.c: At top level:
cc1: error: unrecognized command line option '-Wno-nullability-completeness' [-Werror]
cc1: error: unrecognized command line option '-Wno-missing-variable-declarations' [-Werror]
cc1: error: unrecognized command line option '-Wno-system-headers-Wno-thread-safety' [-Werror]
cc1: all warnings being treated as errors
goes after march-15 release
during bugwash, @bsdphk brought up the question if this could affect vsl clients, and while it does not look like it, git grep
suggests that we should change some other places where FetchError
is referred.
after march15 release
Some scenarios like hitting
send_timeout
because of a dripping write leaves no clue that an error occurred from the client request logs alone.This introduces a client counterpart for:
Refs 49e44e390488c57c03d8111fcd43e245933bc151