xauxag 0 Posted January 5 (edited) Hi, Here may be a bug,would you please have a look?, unit: OverbyteIcsWebSocketCli.pas position: line 1494 ServerKey := Base64Encode(String(SHA1ofStr(s))); ------------ test: x3JJHMbDL1EzLkh9GBhXDw==258EAFA5-E914-47DA-95CA-C5AB0DC85B11 expect: HSmrc0sMlYUkAGmm5OPpG2HaGWk= output:HSmrc0sMlYUkAGmm5OPpG2E/aQ== ------------ test code: procedure TForm1.Button3Click(Sender: TObject); var s : AnsiString; ServerKey:string; begin s := AnsiString(edit1.Text); ServerKey := Base64Encode(String(SHA1ofStr(s))); memolog.Lines.Add(ServerKey); end; node.js - Base64 encoding for Sec-WebSocket-Accept value - Stack Overflow Edited January 5 by xauxag Share this post Link to post
Angus Robertson 574 Posted January 5 Are you suggesting there is a problem with the Base64Encode function creating incorrect output? Have you attempted to debug and fix the problem yourself, or are you expecting someone else to do it? I've been testing the OverbyteIcsWebSocketCli unit this week and it works fine. Angus Share this post Link to post
Virgo 18 Posted January 5 Really strange... That output value decodes to hexencoded value 1d29ab734b0c9585240069a6e4e3e91b613f69 Correct value would be 1d29ab734b0c9585240069a6e4e3e91b61da1969 So da19 to 3f or lead surrogate-da19 to question mark? Some sort weird unicode issue? Share this post Link to post
xauxag 0 Posted January 5 (edited) 3 hours ago, Angus Robertson said: Are you suggesting there is a problem with the Base64Encode function creating incorrect output? Have you attempted to debug and fix the problem yourself, or are you expecting someone else to do it? I've been testing the OverbyteIcsWebSocketCli unit this week and it works fine. Angus Yes I am pretty sure it's a bug, I have debug into ICS and found the two function that caused failure of websocket connection , and am expecting fixing from more official way . I just a little time ago found it and am looking into those code , and not sure exact the correcting till now , so any suggestions are welcome Edited January 5 by xauxag Share this post Link to post
Angus Robertson 574 Posted January 5 I will need to reproduce connection failures before I look into fixing them. So I need a Websocket server to test against. My only suggestion is to try the client against other servers and see if any fail, because no-one else has reported any problems. Angus Share this post Link to post
Kas Ob. 121 Posted January 5 5 hours ago, xauxag said: node.js - Base64 encoding for Sec-WebSocket-Accept value - Stack Overflow I don't see any bug or wrong here, used the same 3 lines like yours and the same value. On other hand if ICS Base64 encoding is broken then it is amazing to perform anything at all ! Share this post Link to post
Kas Ob. 121 Posted January 5 5 hours ago, xauxag said: test code: procedure TForm1.Button3Click(Sender: TObject); var s : AnsiString; ServerKey:string; begin s := AnsiString(edit1.Text); ServerKey := Base64Encode(String(SHA1ofStr(s))); memolog.Lines.Add(ServerKey); end; Please try this instead procedure TForm1.Button3Click(Sender: TObject); var s: AnsiString; ServerKey: string; begin s := AnsiString(edit1.Text); ServerKey := Base64Encode(SHA1ofStr(s)); Memo1.Lines.Add(ServerKey); end; 1 Share this post Link to post
David Heffernan 2345 Posted January 5 I'm just curious but why are you using AnsiString casts which produce behaviour dependent on the executing process locale? Did you mean to encode as UTF8? Share this post Link to post
Angus Robertson 574 Posted January 6 Quote why are you using AnsiString casts The Base64Encode and SHA1ofStr functions were written 20 to 25 years ago before Unicode, we convert to UTF-8 before calling them if needed. In this unit, the input to create the server key is a Base64 encoded random number we created and sent to the server earlier and a GUID so all simple ASCII. Angus Share this post Link to post
Brian Evans 105 Posted January 7 The bug is in your code - not all binary data can cleanly be put in a string directly. The output of the hash function is in TBytes and the encode function can take TBytes as input. That string() cast you inserted between them is causing issues depending on the binary hash value. 3 Share this post Link to post
Angus Robertson 574 Posted January 8 Quote the output of the hash function is in TBytes and the encode function can take TBytes as input. The output of the hash function is AnsiString and the input to encode is AnsiString, so the String cast is unnecessary. I'll remove it for the new version going to SVN today. Angus Share this post Link to post
Angus Robertson 574 Posted January 9 This Websocket client problem should be fixed in SVN and the overnight zip, assuming it was an ANSI/Unicode casting problem with non-English character set conversions, which I'm unable to reproduce. It did not affect Websocket server which had no casting, and yet always worked against our client. The real problem is ICS has overloaded versions of many functions for ANSI and Unicode, and the compiler does not always choose the correct version if input and output parameters don't match, fixed by using a specific ANSI function. I'm slowly adding TBytes versions for binary data to avoid such problems. but it's a long job. Angus 1 Share this post Link to post
Kas Ob. 121 Posted January 9 If @xauxag can provide some more input about the locale and settings then it is possible to test it. I had being using https://github.com/xupefei/Locale-Emulator for years for the purpose of testing when such problem appear or for testing a specific non internationalized (English) application. Share this post Link to post
David Heffernan 2345 Posted January 9 On 1/8/2024 at 8:54 AM, Angus Robertson said: the input to encode is AnsiString It makes no sense to me that a base64 function could ever accept string input. Input base to be bytes. On 1/8/2024 at 8:54 AM, Angus Robertson said: The output of the hash function is AnsiString This is also kinda weird. Shouldn't it be the native type for text, string. 1 Share this post Link to post
Virgo 18 Posted January 10 Since hash function results binary hash and not hexencoded one and the function predates Unicodestring Ansistring result makes sense. UnicodeString for binary hash would be the worst type. Share this post Link to post
David Heffernan 2345 Posted January 10 (edited) 22 minutes ago, Virgo said: UnicodeString for binary hash would be the worst type. I was referring to base64. That takes binary input (so bytes) and outputs text (so string). For sha1 the yes the input and output should be binary, that is bytes. The you'd have helper functions to take hash output and write in friendly hex format. At no point should there be AnsiString. I mean I hope you aren't proposing AnsiString as a place to hold bytes. Edited January 10 by David Heffernan 1 Share this post Link to post
Virgo 18 Posted January 10 2 minutes ago, David Heffernan said: I mean I hope you aren't proposing AnsiString as a place to hold bytes. What I ment was, that I got an impression, that SHA1ofStr and Base64Encode functions were originally written for pre-Unicode Delphi. And to make them work on Unicode Delphi version String was just replaced with AnsiString. And on old Delphi versions it was usual to use string as byte buffer. Event Delphi itself did it. Share this post Link to post
David Heffernan 2345 Posted January 10 29 minutes ago, Virgo said: And on old Delphi versions it was usual to use string as byte buffer. Event Delphi itself did it. On modern delphi you can't get away with this as has been discussed many times. Anyway just because people got things wrong 20 years ago is no reason to continue getting things wrong. Share this post Link to post
Angus Robertson 574 Posted January 10 ICS still supports Delphi 7 and later, so needs AnsiStrings. I am trying to modernise some code, but can not break old functions that people have used in applications for 20 years. I'm expecting some bad feedback as people upgrade to ICS V9.1 which has some non-backward compatible SSL/TLS changes, new units and conceptual changes. SVN notes have all the details. Angus 1 Share this post Link to post
David Heffernan 2345 Posted January 10 2 hours ago, Angus Robertson said: ICS still supports Delphi 7 and later, so needs AnsiStrings. I mean, byte arrays exist in Delphi 7 Share this post Link to post
Angus Robertson 574 Posted January 10 Quote I mean, byte arrays exist in Delphi 7 Correct, and ICS has always used byte arrays in low level functions, the MD5 digest is a byte array, although the Sha1 digest is an AnsiChar array since it was contributed. But there were no TBytes library functions in Delphi 20 years old, so AnsiString was commonly used. I've added a lot of TBytes functions to ICS in the last few years so that support for old compilers has continued, and a TBytes version of Base64 conversion recently. But old code only gets modernised when updated for some reason. Angus Share this post Link to post
David Heffernan 2345 Posted January 10 1 hour ago, Angus Robertson said: But old code only gets modernised when updated for some reason. There's plenty of reasons not to use AnsiString, unless the library is not developed any more. Share this post Link to post
xauxag 0 Posted January 10 On 1/6/2024 at 1:02 AM, Kas Ob. said: Please try this instead procedure TForm1.Button3Click(Sender: TObject); var s: AnsiString; ServerKey: string; begin s := AnsiString(edit1.Text); ServerKey := Base64Encode(SHA1ofStr(s)); Memo1.Lines.Add(ServerKey); end; Thanks, It works 19 hours ago, Kas Ob. said: If @xauxag can provide some more input about the locale and settings then it is possible to test it. I had being using https://github.com/xupefei/Locale-Emulator for years for the purpose of testing when such problem appear or for testing a specific non internationalized (English) application. It is Chinese simplify Share this post Link to post