Hi,

See: https://bugs.freepascal.org/view.php?id=32837

TFloatHelper gives incorrect results for Mantissa and Exponent, also BuildUp fails.
I posted a patch in the bugtracker.

Can I get some feedback from a devel please?

Bart

Patience, please.

I have seen the bugreport, but I need time to check everything. I have assigned the bugreport to myself, so I won't forget.

Michael.

Michael Van Canneyt wrote

Patience, please. I have seen the bugreport, but I need time to check everything.

I do have patience.
It took me quite some time to try and understand how all the different types of floats are stored in memory and how they work, so I don't expect you to apply my patch just on face value.
Also, probably some more comparison against Delphi is needed, which I cannot do (I can't keep asking other people to build and run Delphi programs for me).

But I do think it would be nice if issues with patches would get some feedback. The issue had not been acknowledged, nor confirmed or monitored by any of the devels.
This is rather discouraging for people supplying patches (especially if they have enjoyed the road it took to get there).

Please note that this is not meant as criticism, but as constructive feedback. Also please note that I'm currently writing in a language that is not native to me, so nuances may get lost in translation, so to speak.

I have assigned the bugreport to myself, so I won't forget.

Thanks.

Bart

Bart wrote

Also, probably some more comparison against Delphi is needed, which I cannot do (I can't keep asking other people to build and run Delphi programs for me).

IIRC, the Delphi starter edition is free of charge. It should be sufficient for such simple tests.

Ondrej

Ondrej Pokorny wrote

IIRC, the Delphi starter edition is free of charge. It should be sufficient for such simple tests.

License for use until your individual revenue from Delphi applications or company revenue reaches $1,000 US or your development team expands to more than 5 developers

My development team is the Lazarus development team, so definitely more than 5 people .... ;-)

Does it (D Starter) behave "nice" alongside fpc?

Bart

I don't use Delphi Starter. But I do have (several versions of) Delphi installed alongside with FPC/Lazarus and I don't have a problem. The only thing is that Delphi ships its own Make.exe and adds it to OS PATH, so you need to give absolute path to the FPC make.exe in Lazarus IDE options dialog.

Ondrej

When we are at this topic: could an FPC developer take a look at https://bugs.freepascal.org/view.php?id=31985 ? The issue has been ignored for half a year although:

1.) It has a patch attached.
2.) It's a Delphi compatibilty fix.
3.) Michael stated in
https://bugs.freepascal.org/view.php?id=20166#c76029 that it should work. The patch fixes compiler/streaming behavior to match Michael's statement.
4.) It's important for the Lazarus IDE: string properties with non-empty default strings cannot be correctly stored in the form's LFM.

I don't know what more points the FPC team needs to check an issue :)

Ondrej

I seem to have missed this bugreport. I assigned it to myself, will look at it this weekend.

Michael.

Bart wrote

Does it (D Starter) behave "nice" alongside fpc?

I have the Starter 10.1 and 10.2 running on the same machine which has also several Lazarus/fpc versions. No problems, except for scripts because Delphi merges itself into the system path so that the fpc tools are not found, but those of Delphi.

In our previous episode, Werner Pamler said:

Similarly with e.g. cygwin or mingw or various embedded toolchains (but not Microsoft Visual Studio).

I have a small 10 line generator that generates batchfiles like below, that can be called multiple times, and save the original path, allowing to switch. This one is for FPC, but I have similar ones for cygwin etc.

@echo off

Rem we save oldpath only once, so we can always revert to it.

if "%OLDPATH%" neq "" goto :nosave
set OLDPATH=%PATH%
:nosave
SET PATH=%OLDPATH%
PATH c:\pp32\bin\i386-win32;%PATH%

Michael Van Canneyt wrote

I seem to have missed this bugreport. I assigned it to myself, will look at it this weekend.

Thank you Michael, I highly appreciate your work!

Ondrej