Re: [sqlite] SQLITE_OMIT_BUILTIN_TEST renders fix for ticket [da784137] useless

[Richard Hipp]

On 12/7/16, Ralf Junker  wrote:
>
> I assumed that
> binaries on sqlite.org would omit test code for best performance.

We go by the principle of "Fly what you test, and test what you fly".

>
>> The SQLITE_OMIT_BUILTIN_TEST option is untested and unsupported.
>
> This was not clear to me from the docs:

I agree.  We'll be fixing that.  Sorry for the confusion.

>
> I welcome SQLITE_OMIT_BUILTIN_TEST for smaller binaries and better
> performance.

The SQLITE_OMIT_BUILTIN_TEST compile-time option improves performance
by 0.015% and decreases binary size by 0.65%.

-- 
D. Richard Hipp
d...@sqlite.org
___
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users

Re: [sqlite] SQLITE_OMIT_BUILTIN_TEST renders fix for ticket [da784137] useless

[Ralf Junker]

On 07.12.2016 14:21, Richard Hipp wrote:


Side note: I notice that the SQLite binaries (Windows, at least) are not
compiled with SQLITE_OMIT_BUILTIN_TEST and not affected by the problem.
Is there a reason to omit SQLITE_OMIT_BUILTIN_TEST from the builds, as
it adds at least some overhead?

>

The reason to omit SQLITE_OMIT_BUILTIN_TEST is that it renders SQLite
untestable.


Because of SQLite's well known pre-release tests, I assumed that 
binaries on sqlite.org would omit test code for best performance. After 
all, they are meant for production, not testing, are they not?



The SQLITE_OMIT_BUILTIN_TEST option is untested and unsupported.


This was not clear to me from the docs:

  http://www.sqlite.org/compile.html#omit_builtin_test


There has been a proposal circulating among the developers to disable
SQLITE_OMIT_BUILTIN_TEST.  In other words, make
SQLITE_OMIT_BUILTIN_TEST a no-op.  Your report adds fresh impetus to
that proposal.


I welcome SQLITE_OMIT_BUILTIN_TEST for smaller binaries and better 
performance. Up to now it has not caused me any problems. So many thanks 
for the quick fix:


  http://www.sqlite.org/src/info/afab166313e0b8ad

Ralf
___
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users

Re: [sqlite] SQLITE_OMIT_BUILTIN_TEST renders fix for ticket [da784137] useless

[Richard Hipp]

On 12/7/16, Ralf Junker  wrote:
>
> Side note: I notice that the SQLite binaries (Windows, at least) are not
> compiled with SQLITE_OMIT_BUILTIN_TEST and not affected by the problem.
> Is there a reason to omit SQLITE_OMIT_BUILTIN_TEST from the builds, as
> it adds at least some overhead?

The reason to omit SQLITE_OMIT_BUILTIN_TEST is that it renders SQLite
untestable.

By "untestable" in the previous sentence, I mean it make it impossible
to do 100% MC/DC testing of the delivered binary.  This means, for
example, that we cannot verify that the compiler didn't make an error
while translating the SQLite source code into machine code.

The SQLITE_OMIT_BUILTIN_TEST option is untested and unsupported.

There has been a proposal circulating among the developers to disable
SQLITE_OMIT_BUILTIN_TEST.  In other words, make
SQLITE_OMIT_BUILTIN_TEST a no-op.  Your report adds fresh impetus to
that proposal.

-- 
D. Richard Hipp
d...@sqlite.org
___
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users