Spelt
commented
8 years ago Hi Carlo,
Nice one, weird that one of the Russian committers did not catch this one :-)
Android: Confirmed OK IOS 32 Bit : Confirmed OK IOS 64 Bit : Confirmed OK
Op 30 aug. 2016, om 13:56 heeft Carlo Sirna notifications@github.com het volgende geschreven:
I created a barcode containing utf8 characters taken from both the cyrillic and european charset (to be sure that they can't be all casted to ansi characters without seeing any data loss: your ansi charset can show only one of the two), and I added this barcode to the unit tests.
The test was failing: the string was surely being casted to a 8 bit ansi charset, losing any character that can't be represented in the local 8 bit charset (as I suspected this would happen because there was a compiler warning about this happening). the warining was in in TReadResult.SetText.
The suspiscious code was the call to UTF8ToString(AValue) before setting the FText value.
I put a breakpoint there and I did see that AValue was already containing te correct unicode string with all the characters I was expecting to read from the QR-CODE, so there was no need to do any conversion.
Surely the directly Assigning AValue to FText, without any transformation solved the bug under Win32 and Win64 compilers. I don't know if that call to UTF8ToString was needed for Android/Ios: I have never developed with next gen compilers and I can't test it.
Before applying this pull request could you please try to scan the barcode I added with ZXIng compiled for android/ios? you should get this string: "ру́сский язы́к, russkij jazyk èàòù"
Note: I edited the literal strings in the test cases in order to use escape sequences with unicode values of characters instead of leaving the accented letters in the source: this could give problems to programmers who don't use a west-european windows installation. Another option could be to save the sources in UTF-8 format instead of using ANSI...
You can view, comment on, or merge this pull request online at:
https://github.com/Spelt/ZXing.Delphi/pull/39 https://github.com/Spelt/ZXing.Delphi/pull/39 Commit Summary
added utf8 test for barcode containing mixed european and russian characters. it failed. fixed bug File Changes
M Lib/Classes/Common/ZXing.ReadResult.pas https://github.com/Spelt/ZXing.Delphi/pull/39/files#diff-0 (4) A UnitTest/Images/utf8-test.png https://github.com/Spelt/ZXing.Delphi/pull/39/files#diff-1 (0) M UnitTest/Test.pas https://github.com/Spelt/ZXing.Delphi/pull/39/files#diff-2 (18) Patch Links:
https://github.com/Spelt/ZXing.Delphi/pull/39.patch https://github.com/Spelt/ZXing.Delphi/pull/39.patch https://github.com/Spelt/ZXing.Delphi/pull/39.diff https://github.com/Spelt/ZXing.Delphi/pull/39.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Spelt/ZXing.Delphi/pull/39, or mute the thread https://github.com/notifications/unsubscribe-auth/AMEt1TuBS1L46aB90Ips_xYysZshuiQ6ks5qlBpbgaJpZM4JwcZN.
csm101
I created a barcode containing utf8 characters taken from both the cyrillic and european charset (to be sure that they can't be all casted to ansi characters without seeing any data loss: your ansi charset can show only one of the two), and I added this barcode to the unit tests.
The test was failing: the string was surely being casted to a 8 bit ansi charset, losing any character that can't be represented in the local 8 bit charset (as I suspected this would happen because there was a compiler warning about this happening). the warining was in in TReadResult.SetText.
The suspiscious code was the call to UTF8ToString(AValue) before setting the FText value.
I put a breakpoint there and I did see that AValue was already containing te correct unicode string with all the characters I was expecting to read from the QR-CODE, so there was no need to do any conversion.
Surely the directly Assigning AValue to FText, without any transformation solved the bug under Win32 and Win64 compilers. I don't know if that call to UTF8ToString was needed for Android/Ios: I have never developed with next gen compilers and I can't test it.
Before applying this pull request could you please try to scan the barcode I added with ZXIng compiled for android/ios? you should get this string: "ру́сский язы́к, russkij jazyk èàòù"
Note: I edited the literal strings in the test cases in order to use escape sequences with unicode values of characters instead of leaving the accented letters in the source: this could give problems to programmers who don't use a west-european windows installation. Another option could be to save the sources in UTF-8 format instead of using ANSI...