Today I had problems in the FPDocManager, where the scanner/parser throws an exception "String list does not allow duplicates". Reason may be duplicate include paths, definitely a duplicate -dCPU32 from both the parser and the commandline.

This error now also occurs when the RTL 2.6 docs are build from the commandline (Win7):

make rtl.chk

What happend to the Makefile, or to the PParser/Scanner?

Another parser error occurs in systemh.inc line 283 col. 30:

   UCS4Char            = type 0..$10ffff;

where the parser expects an "=" after "type". Wasn't this syntax recognized already before 2.6?

Can somebody confirm above problems, or are these due to bad modifications on my machine?

DoDi

Hans-Peter Diettrich wrote

Today I had problems in the FPDocManager, where the scanner/parser throws an exception "String list does not allow duplicates". Reason may be duplicate include paths, definitely a duplicate -dCPU32 from both the parser and the commandline.

The sources have been stripped from dupError a long time ago. So I can't reproduce this.

Another parser error occurs in systemh.inc line 283 col. 30: UCS4Char = type 0..$10ffff; where the parser expects an "=" after "type". Wasn't this syntax recognized already before 2.6?

As far as I know, I have fixed this already in trunk.

Michael.

See pscanner.pp line 947, where FDefines is created with dupError.

The reported problems exist since my last FPC 2.7 SVN update (yesterday), they did not exist in the last months.

DoDi

What is the problem ? Defines or include paths ?

If you think there is a problem with double defines, please state this clearly. From your report I understand you think there is a problem with include paths. I cannot reproduce this: Yesterday I built FCL and RTL with ALL include paths doubled. Everything went well, so I think the problem is on your machine.

If you think there is an additional problem with defines, then state this clearly or at the very least in a separate report.

For this one, please file a separate bug report.

Michael.

The problem is at least one list which reject duplicates. Since the error message is too vague, I only can guess what may be the exact problem. Defines definitely are a problem.

> The reported problems exist since my last FPC 2.7 SVN update > (yesterday), they did not exist in the last months. For this one, please file a separate bug report.

I could find out one trouble maker: Windows Update :-( Even if I had disabled all updates, this month 2 updates have been installed - restored to preceding restorepoint and now the Makefile works again :-)

I'm still clueless about the syntax errors. They don't show up when calling fpdoc on the commandline, but they show up when using the fpdoc units in FPDocManager. I'll have to check all my installations now, which may be garbled from preceding attempts to fix the problems caused by Windows.

DoDi

Not here.

There is only 1 dupError in all the sources of fcl-passrc:

home: >grep -i dupError *

pscanner.pp:580:  FStreams.Duplicates:=dupError;
And that one is on purpose. In all other code there is no dupError.

In fpdoc there is no dupError at all:

home: >grep -ni dupError *.pp
home: >

All the code is checked in, so I suggest you check your installation.

Michael.

Perhaps not on your platform, but perhaps on many others. The -dCPU32 in the fpdoc commandline may be inserted on a Win64 platform only...

There is only 1 dupError in all the sources of fcl-passrc:

Yes, see above where it sits.

DoDi

Look, I tested by inserting the same define twice on the command-line, and all worked fine. Further, there is not a single piece of evidence that dupError is used in defines in the SVN sources.

> > There is only 1 dupError in all the sources of fcl-passrc: Yes, see above where it sits.

Indeed. It is not in the defines. In the TStreamResolver, and this class is currently only used in the testcases.

I'm sorry, but the problem is on your pc, not in FPC.

Michael.

Always doubt yourself.

I checked here at work, and here the SVN copy is different I committed a fix, please test rev. 20169.

Now I should learn to doubt and check before writing an answer.

Michael.

I checked here at work, and here the SVN copy is different than at home. So I committed a fix, please test rev. 20169.

Just tested: the dupes error has gone away :-)

Now remain the syntax errors, see Mantis #21148.

DoDi

Great.

Now remain the syntax errors, see Mantis #21148.

For those, I still need to test.

BTW. In future, please file separate reports for separate issues.

There are 2 distinct problems:
* the duplicate defines,
and
* the parsing of certain constructs.

Michael.

BTW. In future, please file separate reports for separate issues. There are 2 distinct problems: * the duplicate defines, and * the parsing of certain constructs.

Both kinds of errors appeared at the same time, suggesting to me that they have the same origin. Please forgive me that I couldn't spot the real bugs more precisely :-(

DoDi