Difference between revisions of "Issues 3.2.0"

From Free Pascal wiki
Jump to navigationJump to search
(→‎ISO mode: Fixed)
 
(5 intermediate revisions by 3 users not shown)
Line 16: Line 16:
 
=== Other documentation issues ===
 
=== Other documentation issues ===
  
* Runtime error 213 trying to use the textmode IDE help index. {{MantisLink|37094}}
+
* <s>Runtime error 213 trying to use the textmode IDE help index. {{MantisLink|37094}}</s> Fixed in revisions 45406,45407.
 
* The Language Reference, Programmer's Guide, and User's Guide CHM files have an odd string on their first page: "mei 2019". Additionally the accented letters in author names aren't displayed correctly in those files. The FPDoc CHM has a similar front page, but it does looks correct.
 
* The Language Reference, Programmer's Guide, and User's Guide CHM files have an odd string on their first page: "mei 2019". Additionally the accented letters in author names aren't displayed correctly in those files. The FPDoc CHM has a similar front page, but it does looks correct.
  
Line 23: Line 23:
 
* The Intel installer currently installs even if the command line utilities are not yet installed. This results in missing parameters in the generated /etc/fpc.cfg (even if they are installed afterwards; you have to reinstall fpc to regenerate fpc.cfg)
 
* The Intel installer currently installs even if the command line utilities are not yet installed. This results in missing parameters in the generated /etc/fpc.cfg (even if they are installed afterwards; you have to reinstall fpc to regenerate fpc.cfg)
  
* The Intel installer currently installs the man pages into /usr/local/man/[man1|man5] where they are not found. They should be installed into /usr/local/share/man/[man1|man5]. I reported here for 3.0.4: https://bugs.freepascal.org/view.php?id=35969
+
* <s>The Intel installer currently installs the man pages into /usr/local/man/[man1|man5] where they are not found. They should be installed into /usr/local/share/man/[man1|man5]. I reported here for 3.0.4: https://bugs.freepascal.org/view.php?id=35969</s>
 
** Fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
 
** Fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
* A man page is installed for fp (text mode IDE), but there is no fp binary installed. {{MantisLink|36861}}
+
* <s>A man page is installed for fp (text mode IDE), but there is no fp binary installed. {{MantisLink|36861}}</s>
 
** Fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
 
** Fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
* Test suite (make full TEST_FPC=/usr/local/bin/ppcx64) fails to complete with an error. See https://sentinel.sentry.org/mylog for the log file.
+
* <s>Test suite (make full TEST_FPC=/usr/local/bin/ppcx64) fails to complete with an error. See https://sentinel.sentry.org/mylog for the log file.</s>
 
** That's because a merge was missing, fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
 
** That's because a merge was missing, fixed. [[User:Jonas|Jonas]] ([[User talk:Jonas|talk]])
  
Line 68: Line 68:
 
* Does not ship pas2js binary, but does ship a pas2js units directory?
 
* Does not ship pas2js binary, but does ship a pas2js units directory?
 
* [https://forum.lazarus.freepascal.org/index.php/topic,49110.msg362517.html#msg362517 Forum post] Val procedure (and StrToFloat as well) fails with SIGFPE with Math.MaxDouble.
 
* [https://forum.lazarus.freepascal.org/index.php/topic,49110.msg362517.html#msg362517 Forum post] Val procedure (and StrToFloat as well) fails with SIGFPE with Math.MaxDouble.
* $Unitpath directive doesn't work correctly. {{MantisLink|37095}}
+
* <s>$Unitpath directive doesn't work correctly. {{MantisLink|37095}}</s> Fixed in 45410, may or may not actually make it to 3.2.0.
 
* An initial SetLength on a local managed array variable now causes a warning of uninitialized usage even though managed variables are always automatically initialized. Easy to work around but a nuisance, maybe should be mentioned in release notes if behavior is not changed. {{MantisLink|34169}}
 
* An initial SetLength on a local managed array variable now causes a warning of uninitialized usage even though managed variables are always automatically initialized. Easy to work around but a nuisance, maybe should be mentioned in release notes if behavior is not changed. {{MantisLink|34169}}
 +
* <s>Can no longer get the address of an index of a function result string, without saving the result string into a variable first. Could be intentional. {{MantisLink|37098}}</s> This should never have worked in the first place, so it's more correct now.
  
 
=== ISO mode ===
 
=== ISO mode ===
<s>* Assigning a character to a file variable leads to a runtimes error 104, whereas trunk works. See https://bugs.freepascal.org/view.php?id=36902. </s> This has been fixed with commit 44301 to trunk and I expect that RC2 will have it.
+
* <s>Assigning a character to a file variable leads to a runtimes error 104, whereas trunk works. See https://bugs.freepascal.org/view.php?id=36902. This has been fixed with commit 44301 to trunk and I expect that RC2 will have it.</s> Fixed [[User:Mischi|Mischi]] ([[User talk:Mischi|talk]])
  
 
== Merge requests ==
 
== Merge requests ==
  
 
* r44127 needs to be merged to fix a Windows Registry {{MantisLink|36663}}.
 
* r44127 needs to be merged to fix a Windows Registry {{MantisLink|36663}}.
 +
* r44179 should be merged: New Property EmptyImputRaisesError to avoid TRegExpr Error "empty input string" {{MantisLink|36603}}
  
 
== Post RC1 merges ==
 
== Post RC1 merges ==

Latest revision as of 14:41, 24 May 2020

Issues with 3.2.0 release candidates

This page is supposed to contain all issues discovered while testing release candidates and their status. Make sure to include information about the platform where the problem appears (if applicable). Once the issue gets resolved, you should strike it over and add number of SVN revision fixing it. If the revision number refers to a SVN repository different from "fpc" (i.e. "fpcbuild" or "fpcdocs"), include information about the SVN repository too.

Issues with 3.2.0-rc1

faq.txt

readme.txt typos

  • "Both these adresses are for mailing lists. If you're not subscribed," -- adresses should be addresses.
  • "order to make the compiler/rtl & IDE in a resonable time (eg <30 minutes)" -- resonable should be reasonable.

Other documentation issues

  • Runtime error 213 trying to use the textmode IDE help index. Issue #37094 Fixed in revisions 45406,45407.
  • The Language Reference, Programmer's Guide, and User's Guide CHM files have an odd string on their first page: "mei 2019". Additionally the accented letters in author names aren't displayed correctly in those files. The FPDoc CHM has a similar front page, but it does looks correct.

macOS

  • The Intel installer currently installs even if the command line utilities are not yet installed. This results in missing parameters in the generated /etc/fpc.cfg (even if they are installed afterwards; you have to reinstall fpc to regenerate fpc.cfg)
  • The file "Getting Started (Intel).rtf": It contains numerous references to very old details, which makes it long and hard to check, what is relevant to recent systems. I would replace most of that by a link to an according wiki page. It also says, that you cannot compile 64 bit GUI programs with Lazarus, without mentioning the cocoa, Qt4 and Qt5 64 bit variants of Lazarus. Please note that this text is also displayed at the start of the installation.
    • The details on how to use FPC with older (Mac) OS X/macOS versions will remain there as long as those OSes remain supported by FPC. I will update the Lazarus part. Jonas (talk)
      • In order to get rid of the overly dull and outdated impression, I suggest to reorder the points in better clarity with general ones first, followed by system version specific ones, from newest to oldest systems.
  • The file "readme.txt"
    • It still has 3.0.0 as version number at several places.
    • It mentions the GNU debugger and not lldb
    • Many feature are not really relevant for macOS or even not available.
    • The QuickStart has some minor outdated stuff.
      • The readme file has not yet been updated for FPC 3.2, and is the same in all platform installers. macOS-specific things are in the Getting Started files. Jonas (talk)
  • The file "whatsnew.txt"
    • News about 3.2.0 are missing
      • This file has not yet been updated either. Jonas (talk)
  • not really a bug of fpc, but still a noteworthy issue: lazarus-cocoa does not build
    • fpc 3.2.0 comes with updated cocoa headers, which breaks building lazarus-cocoa. Bugtracker: https://bugs.freepascal.org/view.php?id=36864 (resolved with lazarus commits 62885 (trunk) and 62901 (fixies_2_0). Thanks to the Lazarus team.)
      • That was because of a missing merge to 3.2.0. Those changes to Lazarus should be reverted again. Jonas (talk)

Linux

As tested on Ubuntu20.04 beta but not specific.

Testing instructions says run every binary. I am not familiar with all of them so I ran some with just '-h' as a parameter. All do have correct binary format but not all will accept that '-h', of those, most examined my '-h' and rejected it as an invalid parameter, thats OK. However -

  • unitdiff - does not support -h, crashes with bad input. OK with good input.
  • mkx86ins - crashes with bad input.
  • The readme.txt, faq.txt and whatsnew.txt all refer to previous version being 3.0.0 not 3.0.4. Maybe this is as intended but, for example, the whatsnew does have specific content for 2.6.0, 2.6.2 and 2.6.4.

Go32v2

  • The Go32v2 textmode IDE has no GDB due to builidng issues, reported by Pierre while uploading. Try to find old known working one?

Win32 and possibly others

  • Does not ship pas2js binary, but does ship a pas2js units directory?
  • Forum post Val procedure (and StrToFloat as well) fails with SIGFPE with Math.MaxDouble.
  • $Unitpath directive doesn't work correctly. Issue #37095 Fixed in 45410, may or may not actually make it to 3.2.0.
  • An initial SetLength on a local managed array variable now causes a warning of uninitialized usage even though managed variables are always automatically initialized. Easy to work around but a nuisance, maybe should be mentioned in release notes if behavior is not changed. Issue #34169
  • Can no longer get the address of an index of a function result string, without saving the result string into a variable first. Could be intentional. Issue #37098 This should never have worked in the first place, so it's more correct now.

ISO mode

Merge requests

  • r44127 needs to be merged to fix a Windows Registry Issue #36663.
  • r44179 should be merged: New Property EmptyImputRaisesError to avoid TRegExpr Error "empty input string" Issue #36603

Post RC1 merges