Community
Participate
Working Groups
The Ant log file uses the Linux carriage return as opposed to the system carriage return character. see https://bugs.eclipse.org/bugs/show_bug.cgi?id=54042 for my ant attachment that has the Linux style carriage returns.
Thanks for the bug report but this is not the Ant log file but rather the compiler log file generated by using the javac task and the jdtCompilerAdapter via the Export operation. I am not sure how you could have gotten the wrong carriage returns...moving to JDT Core for comment.
Wassim, have you seen anything like this? Not sure where this bug should live.
Logically, it belongs in JDT/Core because the log files are generated during the javac task, as Darins points out. I'm not sure how these files are written to, but, if each line is terminated with System.getProperty("line.separator"), then we would get a platform- specific carriage return.
All lines written in the log use the println(...) methods of the PrintWriter class. We never explicitely use the line delimiter. I will investigate further.
Olivier, PrintWriter.println() is sufficient. It sends values to the output stream followed by an end of line character. The end of line sequence used is the default for the system as determined by System.getProperty("line.separator"). For Windows systems,the end of line character is \r\n, for Unix systems, \n, and for Mac systems, \r.
That was my understanding. This is why I am surprised this is an issue. I tried on Windows to generate a log file using the batch compiler and I got Windows line delimiter in it. No problem. It works as expected. Could you please attach in a zip file a log file generated on Windows that contains Linux line delimiter? Maybe your attachment has been converted by Bugzilla.
The ant file was generated on Windows 2000 using JDK 1.4.1_07. I right clicked my fragment.xml file and selected Create Ant Build file. Then I right clicked the build.xml ant file and selected run ant... I selected "build jars" and "gather logs." I then received the compile failed error message as detailed in my other bug report. The I opened the log file. It appears that the log is copying text from the source file. Just before the copied portion there is a funny char, and just after the copied portion there is one. I use the term "one" loosely because I believe there are a few chars there I do not see. here is one example ---------- 6. ERROR in C:\eclipse2.1.2\workspace\WAD_DS_AccessDB\src\com\rigidsoftware\WAD\datasources\access\DBArchitecture.java (at line 150) assert nameSuggestion != null : "System name must not be null."; ^^ Syntax error on token "!=", ";" expected ---------- Just before the spaces in front of the keyword assert. Also just after the semicolon
I submit to you gentlemen, that the errors listed here are not the work of javac. Clearly javac understands the keyword 'assert.' Furthermore, the class files are generated without issue. Therefore, the errors are the result of some other parsing of the class file. A parsing done by an entity that does not comprehend the keyword 'assert.'
Created attachment 8437 [details] Unmolested zip of log zipped to preserve the line characters that bugzilla may be removing.
Version is 2.1.2
Has anyone verified this behavior? I dropped the priority a notch as its just an annoyance. I can use wordpad if I really want to see it correctly.
I found the problem. I have a fix, but we need to update some tests first. I'll see if I can do it today for next integration build. It will be fixed in 3.0 only.
great!
Fixed and released in HEAD. All tests have been updated and passed.
Verified for 3.0 using build I200403250010.
