kicad-developers team mailing list archive

[Wayne Stambaugh <stambaughw@xxxxxxxxx>]

On 7/25/2018 11:15 AM, Mark Roszko wrote:
> @Wayne, The hold up on Windows was that Nick went on vacation
> basically the day after posting the first 5.0 build. So bad timing.>
> 
>> I've changed it to Release for the time being (but someone still needs to accept my pull request in GitHub)
> 
> Unforunately Nick is the only one with privs on the winbuilder repo on
> github unless someone else does from the organization level access
> list.

I guess we will have to wait until Nick returns from vacation to get
this resolved.  Once he gets back, I'll ping him about having him assign
privileges to someone else as a backup so we don't get bit by this in
the future.

> 
>> RelWithDebInfo should generate the same code as Release, and after stripping debug information, the binaries should be identical.
> 
> You know, looking at your commit, its interesting. Because PKGBUILD
> which you change from RelWithDebInfo to Release explicitly was always
> fine, the nightlies never gave that assert (I checked).
> But PKGBUILD-STABLE had no config specified....so perhaps it drifted
> to DEBUG somehow?
> 
> On Wed, Jul 25, 2018 at 9:09 AM Simon Richter <Simon.Richter@xxxxxxxxxx> wrote:
>>
>> Hi,
>>
>> On 25.07.2018 14:16, Wayne Stambaugh wrote:
>>
>>> We should have created a release build of the stable version.  I'm fine
>>> with nightly builds having debugging information.  Stable releases
>>> should not have debug info.
>>
>> These are built as RelWithDebInfo, and the debug information is stripped
>> out during installation, or rather should be (we're trying to keep the
>> MSYS build as close to a standard Linux/BSD build as possible, and they
>> usually archive the debug information separately before stripping, which
>> is why this mode is useful at all).
>>
>> I've changed it to Release for the time being (but someone still needs
>> to accept my pull request in GitHub), but this is indicative of a bug in
>> the build scripts. RelWithDebInfo should generate the same code as
>> Release, and after stripping debug information, the binaries should be
>> identical.
>>
>> There are two problems I see:
>>
>>  - we check explicitly if the build type is Release, which then doesn't
>> match
>>  - we have a redundant -DDEBUG which is explicitly set — release builds
>> have -DNDEBUG, which is set by CMake already, and this is what should
>> switch debugging facilities like asserts.
>>
>> I've submitted a patch to get rid of -DDEBUG two years ago, I doubt it
>> still applies. I can make a new one if it has a chance of being applied.
>>
>>    Simon
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~kicad-developers
>> Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~kicad-developers
>> More help   : https://help.launchpad.net/ListHelp
> 
> 
>