+ From:
+ Behalf Of Stefan. Hille
+ Sent: dinsdag 21 mei 2002 1:33
+
+ Hi
+
+ On Mon, May 20, 2002 at 11:53:54AM +0200, Mattias Gaertner wrote:
+ > On Mon, 20 May 2002 11:36:48 +0200 Mattias Gaertner
+ > wrote:
+ >
+ > I just checked. You are right. They are all beta and all of them do not
+ > contain the patch. Mysterious. I have readded compiler directives. This
+ > time for 1.0.7.
+ > Probably we must once again start building our own compiler. +
+ Don't think we should do this until there are must-have features missing.
+ IMHO this one is only a nice-to-have feature and the ifdef solution should
+ be ok as long as we're far away from 1.0. Building our own compiler
+ would require resources (build, maintain, *document*) we can better spend
+ on our own code.

Building our own here is nothing more that use a stable FPC compiler and add the patches which are not in it. Like the patches Mattias is missing, I've the idea that the fix for #1888 isn't included as well. That one is a really pain if you've to work around it.
Something about #1888
Symptoms are compile and everything seems to work OK, change just a bit so some unit gets recompiled or things are linked differently -> Form streaming produces an AV, change something else and it works probably. One can work around this by publishing most used properties (left, top, height, width, caption etc) again in a derived form class.

If in the final FPC 1.0.6 those patches are missing, then I vote for using a stable 1.0.7

Marc

Hi Mattias,

seems you've updated control.inc a few minutes/hours ago. I get compiler errors now. EG:

  control.inc(812,33) Error: Identifier not found SSTRIPLE

Where is this constant defined? Does it belong to the runtime libs? I'm using

Free Pascal Compiler version 1.0.6-beta [2002/04/30] for i386

Don't know why it's called beta because I downloaded after their official release.

bye,

Stefan

Hi

If bug #1888 is still in 1.0.6 I also vote for using another compiler cause it's a bug which makes reliable development more or less impossible.

bye

It is defined in fpc/fcl/inc/classesh.inc since the fpc 1.0.6 (03/28). Therefore there were some compiler directives {$IFDEF VER1_0_6}. I deleted them, because laz now officially depends on 1.0.6. So, I'm somewhat confused that you were able to compile laz before.

Try another mirror.

Mattias

I just checked. You are right. They are all beta and all of them do not contain the patch. Mysterious. I have readded compiler directives. This time for 1.0.7. Probably we must once again start building our own compiler.

Mattias

Hi

Don't think we should do this until there are must-have features missing. IMHO this one is only a nice-to-have feature and the ifdef solution should be ok as long as we're far away from 1.0. Building our own compiler would require resources (build, maintain, *document*) we can better spend on our own code.

BTW: For those not so familar with cvs:

When I hit the compile error I still wanted to test some things. To make lazarus compile again I checked out the latest working version of control.inc (version 1.49 in this case). This could easily be done by issuing

cvs update -r 1.49 control.inc

in the lcl/include directory. control.inc was now "sticky" and excluded from any furher update. When I saw Mattias mail about the workaround I removed the sticky-flag with

cvs update -A control.inc

bye,

Stefan