recover from doctest timings cache corruption

[vbraun]

Caches can be corrupted as you doctest stuff, you never know when things die. The new doctest framework does not recover gracefully:

[vbraun@volker-desktop hg]$ echo foobar > ~/.sage/timings_dt_test.json 
[vbraun@volker-desktop hg]$ sage -tp sage/doctest/
Running doctests with ID 2013-03-28-12-29-46-c1280e16.
Doctesting 9 files using 10 threads.
sage -t sage/doctest/reporting.py
**********************************************************************
File "sage/doctest/reporting.py", line 103, in sage.doctest.reporting.DocTestReporter.__init__
Failed example:
    DC = DocTestController(DocTestDefaults(),[filename])
Expected nothing
Got:
    Error loading stats from /home/vbraun/.sage/timings_dt_test.json
**********************************************************************
File "sage/doctest/reporting.py", line 130, in sage.doctest.reporting.DocTestReporter.report_head
Failed example:
    DC = DocTestController(DD, [filename])
Expected nothing
Got:
    Error loading stats from /home/vbraun/.sage/timings_dt_test.json

Depends on #14292

CC: @roed314

Component: doctest framework

Reviewer: Volker Braun

Issue created by migration from https://trac.sagemath.org/ticket/14374

[jdemeyer]

comment:1

Volker, #14292 (I have a patch but it's on sage.math which is currently offline) should ensure that the JSON file cannot be corrupted anymore. So, if we get an error loading the JSON file, it should be a genuine error and we want to get a doctest failure.

So we could close this one as duplicate of #14292.

[jdemeyer]

Dependencies: #14292

[vbraun]

comment:2

There is also the issue of changing JSON fields (try pushing/popping #13278, this is the problem that I originally had).

[jdemeyer]

comment:3

Replying to @vbraun:

There is also the issue of changing JSON fields (try pushing/popping #13278, this is the problem that I originally had).

Does it cause the JSON stats to fail to load?

[vbraun]

comment:5

Yes, all doctests failed for me after first pushing #13278

[jdemeyer]

comment:6

Strange, I haven't seen this error while working on #13278. But I'll test it now.

[jdemeyer]

comment:7

I cannot reproduce the problem with #13278.

[jdemeyer]

comment:8

Do we agree that this is fixed by #14292 and that it can be closed?

[vbraun]

Author: Jeroen Demeyer

[vbraun]

Reviewer: Volker Braun

[vbraun]

comment:9

I guess its good enough. Somebody will fall into that trap but at least I'll know how to fix it on my machine.

[jdemeyer]

comment:10

Replying to @vbraun:

Somebody will fall into that trap

If they do, they might have found a genuine error in Sage.

[jdemeyer]

Changed author from Jeroen Demeyer to none