Jar manifest and EOF

[Weijun Wang]

Hi Philipp,

Although I also think this feature might be useless, I feel hesitated to just remove it. How about implement it and let it stay useless? I assume this will not break any existing test or app if no one is actually using the character. Hopefully it's easy to implement.

On the other hand, if there is really a EOF char at the end, there will be a behavior change. If that EOF is at the end an attribute line, that line is ignored now, and it should be visible after the change and this looks like a good thing. If that EOF is after a newline, it will be useless anyway.

Also, my understanding is that hash/signature in a signed jar file are calculated on raw bytes and old signed jars can still be verified. We should take care to use the original raw bytes and not the result of Manifest::write, but I think your previous fixes in this area had already handled this correctly.

Just my $0.02.

Thanks,
Max

> On Oct 11, 2019, at 2:11 PM, Philipp Kunz <philipp.kunz at paratix.ch> wrote:
> 
> Hi,
> 
> The Jar File Specification [1] states that,
> 
>> If the last character of the file is an EOF character (code 26), the
> EOF is treated as whitespace. Two newlines are appended (one for
> editors that don't put a newline at the end of the last line, and one
> so that the grammar doesn't have to special-case the last entry, which
> may not have a blank line after it).
> 
> But I can't see that this is actually the case. See the attached simple
> test that indicates that this statement is not implemented.
> 
> See also potentially remotely related bugs [2] and [3] which both refer
> in my opinion to EOF as the end of the byte stream as opposed to the
> EOF character 26.
> 
> After nobody has fixed this point inside of many years I tend to assume
> that it is not at all necessary or useful. However, I cannot understand
> the intention behind it or how that statement got there in the first
> place and hence might not be aware of the full context.
> 
> Considering the activity about the last line of a manifest not
> terminated by a line break resulting in that line ignored unexpectedly 
> (see [3] through [14]) I doubt that it is promising to attempt to
> implement the EOF according to the specs.
> 
> Instead I'd rather like to propose to update the specification by
> removing the quoted statement.
> 
> I'm not sure if the specification [1] has changed since 13 or where the
> most up to date one is.
> 
> Regards,
> Philipp
> 
> 
> 
> [1] 
> https://docs.oracle.com/en/java/javase/13/docs/specs/jar/jar.html#notes-on-manifest-and-signature-files
> 
> EOF-related:
> [2] https://bugs.openjdk.java.net/browse/JDK-7148584
> [3] https://bugs.openjdk.java.net/browse/JDK-4639129
> 
> Last line break required or missing related:
> [4] https://bugs.openjdk.java.net/browse/JDK-4274235
> [5] https://bugs.openjdk.java.net/browse/JDK-4894998
> [6] https://bugs.openjdk.java.net/browse/JDK-4333854
> [7] https://bugs.openjdk.java.net/browse/JDK-6594305
> [8] 
> http://mail.openjdk.java.net/pipermail/core-libs-dev/2019-February/058774.html
> [9] https://bugs.openjdk.java.net/browse/JDK-4489716
> [10] https://bugs.openjdk.java.net/browse/JDK-4639129
> [11] https://bugs.openjdk.java.net/browse/JDK-4625822
> [12] 
> https://stackoverflow.com/questions/21859417/header-must-be-terminated-by-a-line-break-eclipse-manifest-mf
> [13] https://bugs.eclipse.org/bugs/show_bug.cgi?id=377249
> [14] https://bugs.java.com/bugdatabase/view_bug.do?bug_id=4274235
> 
> <20191011-manifest-eof.patch>