RE: smokecurrent.log vs the report?

Dominic Dunlop wrote:
> On 2008-06-03, at 23:14, H.Merijn Brand wrote:
>=20
>> On Tue, 03 Jun 2008 18:29:46 +0200, Dominic Dunlop
>> <[email protected]> wrote:
>>=20
>>> On 2008-06-03, at 15:17, H.Merijn Brand wrote:
>>>=20
>>>> I think the '?'s should be either 'F' or 'X' here. No idea as yet
>>>> why they are '?'
>>>=20
>>>=20
>>> You can specify a maximum wallclock time for a smoke test. '?' is
>>> the reported status for configurations that did not finish or did
>>> not get run because the smoke was still running when the guillotine
>>> descended. Is that what happened during the test in question?
>>>=20
>>> See "killtime" on configsmoke.3 man page.
>>=20
>> But that would leave you with *one* '?' in the report. I've had many
>> of those. Bram's report had two '?'s
>=20
>=20
> Good point. A bit of trawling shows Steve Hay was getting (nothing
> but) consecutive '?'s in early January reports from Test::Smoke
> v1.20.07 build 1123; he upgraded to Test::Smoke v1.29_60 build 1155,
> and the problem went away. However, as Bram's report is from
> Test::Smoke v1.32 build 1176, that's probably not the answer -- unless
> there was a problem that got fixed and  then regressed.
>=20
> See
>
<http://www.nntp.perl.org/group/perl.daily-build.reports/2008/01/msg5288
8.html
>  >,
>
<http://www.nntp.perl.org/group/perl.daily-build.reports/2008/01/msg5291
0.html
> >, abd
>
<http://www.nntp.perl.org/group/perl.daily-build.reports/2008/01/msg5293
8.html
> >. Steve, can you throw any light on what was going wrong?  =20

Sorry, no idea. The problem I had in January was simply that I was
running a version of Test-Smoke that pre-dated changes to take account
of the new format of Test-Harness output, which resulted (as you've
seen) in *everything* being a '?'. As soon as I upgraded they were all
fixed. I don't recall having had smokes that mostly worked but had a
couple of '?'s.