| RE: [cdt-dev] Internal builder status |
Yes, all very good points.
I don't think you can really get rid of the index-based dependency calculator though. Not all compilers support outputting dependency information, and some sort of solution, however imperfect, is still needed for those situations.
Just to be clear, the first build the user does (FULL_BUILD) will always build everything. Dependency information only comes into play for incremental builds. So the first build will always happen regardless of whether the index is complete or not. The problems occur after that.
Supporting compiler generated dependencies would be worth while, but would take some work. As you say, you'd have to parse the .d files to determine dependencies. You can't just include the .d files into the makefile the way the generated makefile builder does, as having to use make defeats the purpose of the internal builder. The generated makefile builder has it easy that way.
===========================
Chris Recoskie
Team Lead, IBM CDT and RDT
IBM Toronto
"Schorn, Markus" ---06/25/2010 04:07:08 AM---Hi, it is not a good idea to base the builder on the dependencies as found
From: | "Schorn, Markus" <Markus.Schorn@xxxxxxxxxxxxx> |
To: | "CDT General developers list." <cdt-dev@xxxxxxxxxxx> |
Date: | 06/25/2010 04:07 AM |
Subject: | RE: [cdt-dev] Internal builder status |
Sent by: | cdt-dev-bounces@xxxxxxxxxxx |
On Jun 24, 2010 3:12 PM, "Chris Recoskie" <recoskie@xxxxxxxxxx> wrote:
>
> I believe there is an index-based dependency calculator which the internal
> builder uses. This has some flaws though as well, because it hinges on the
> index contents having been generated for the configuration that you are
> building. I think (correct me if I'm wrong) the indexer uses one static
> build configuration by default, so if you switch build configuration from
> say Debug to Release, it might not match what you're building.
>
> In the Debug vs Release case, in all likelihood, your code won't be
> including different headers, so it should work fine for that case 99% of
> the time. The bigger gotcha is if you have multiple configurations that
> target different platforms. Then the set of headers used by each
> configuration is more likely to diverge.
>
> ===========================
> Chris Recoskie
> Team Lead, IBM CDT and RDT
> IBM Toronto
>
>
> |------------>
> | From: |
> |------------>
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |James Blackburn <jamesblackburn@xxxxxxxxx> |
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |------------>
> | To: |
> |------------>
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |"CDT General developers list." <cdt-dev@xxxxxxxxxxx> |
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |------------>
> | Date: |
> |------------>
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |06/24/2010 02:51 PM |
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |------------>
> | Subject: |
> |------------>
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |Re: [cdt-dev] Internal builder status |
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |------------>
> | Sent by: |
> |------------>
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
> |cdt-dev-bounces@xxxxxxxxxxx |
> >--------------------------------------------------------------------------------------------------------------------------------------------------|
>
>
>
>
>
> On 24 June 2010 19:11, Doug Schaefer <cdtdoug@xxxxxxxxx> wrote:
>> On Thu, Jun 24, 2010 at 2:08 PM, James Blackburn
>> <jamesblackburn@xxxxxxxxx> wrote:
>>> On 24 June 2010 18:56, Doug Schaefer <cdtdoug@xxxxxxxxx> wrote:
>>>> I certainly agree with that. If we had someone in there fixing these
>>>> bugs, which as you mention have been there since 6.0 and I think maybe
>>>> longer, then I'd have a different opinion, at least for the cases
>>>> Chris mentions where there is no pre-existing build files and where
>>>> make is not readily available (which isn't that many platforms any
>>>> more).
>>>
>>> I think Chris is on the ball in that this is somewhere we want to be
>>> eventually. As the build model feeds both the makefile generator and
>>> the internal builder, I've been focussing on getting the makefile
>>> builder right for my users. As it stands there are some thorny
>>> platform build integration issues we need to overcome to make it
>>> 'seamless'.
>>
>> James correct me if I'm wrong, but the build model does not generate
>> the build dependencies does it? I thought that came from the gcc
>> compiler itself. At least it did a long time ago. It's the build
>> dependencies that I worry about and what would lead to most build
>> quality issues. If I change a header file, I want a guarantee that all
>> the source files that read that header file get rebuilt.
>
> Yes, you're right. I guess as a fall-back the the internal builder
> could generate and use the dependency list as make does.
> _______________________________________________
> cdt-dev mailing list
> cdt-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/cdt-dev
>
_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev
