Re: Eclipse warning concerning serializable classes
Il giorno 19/nov/04, alle 23:18, Leszek Gawron ha scritto: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; Wouldn't it be better to use the serialver command to generate a serialVersionUID? Ugo -- Ugo Cei - http://beblogging.com/ smime.p7s Description: S/MIME cryptographic signature
Re: Eclipse warning concerning serializable classes
Ugo Cei wrote: Il giorno 19/nov/04, alle 23:18, Leszek Gawron ha scritto: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; Wouldn't it be better to use the serialver command to generate a serialVersionUID? I've found some neat ant task: http://serialver.sourceforge.net/serialver.html -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65
Re: Eclipse warning concerning serializable classes
Il giorno 20/nov/04, alle 10:34, Leszek Gawron ha scritto: I've found some neat ant task: http://serialver.sourceforge.net/serialver.html Then go for it! I hate those warnings and was planning to remove some of them myself, like unread local variables or declared exceptions that are never thrown. Ugo -- Ugo Cei - http://beblogging.com/ smime.p7s Description: S/MIME cryptographic signature
Re: Eclipse warning concerning serializable classes
Ugo Cei wrote:
Il giorno 20/nov/04, alle 10:34, Leszek Gawron ha scritto:
I've found some neat ant task:
http://serialver.sourceforge.net/serialver.html
Then go for it! I hate those warnings and was planning to remove some
of them myself, like unread local variables or declared exceptions
that are never thrown.
Mmmh... before putting serialVersionUID all over the place just because
Eclipse barfs, shouldn't we think a bit more about what fixing this
number actually means?
AFAIK, we need to specify this when a serializable class is subject to
evolutions and we want to ensure that serialized instances of an old
version of a class can be reloaded by a newer version of that same
class. If that constant isn't present, then the JVM will compute one
automatically from the class' structure.
So, if we fix these UIDs, that means that from there on we have to take
great care that future versions of these classes are
structure-compatible with older versions, i.e. no field renaming or
deletion, and no new field whose value is set up in the constructor.
So my feeling is that this could cause more harm than good (having a
serialization exception is better than some other obscure failure)
furthermore considering that Cocoon is not a system where serialized
objects are supposed to be long-lived. And in the hypothetical case
where that would be the case, then the relevant classes, and only these
ones, could have a serialVersionUID, with the appropriate warning in the
javadoc about the needed compatibility of future versions.
So my advice: disable the warning in Eclipse, and that's all!
Sylvain
--
Sylvain Wallez Anyware Technologies
http://www.apache.org/~sylvain http://www.anyware-tech.com
{ XML, Java, Cocoon, OpenSource }*{ Training, Consulting, Projects }
Re: Eclipse warning concerning serializable classes
Sylvain Wallez wrote: Ugo Cei wrote: Il giorno 20/nov/04, alle 10:34, Leszek Gawron ha scritto: I've found some neat ant task: http://serialver.sourceforge.net/serialver.html Then go for it! I hate those warnings and was planning to remove some of them myself, like unread local variables or declared exceptions that are never thrown. Mmmh... before putting serialVersionUID all over the place just because Eclipse barfs, shouldn't we think a bit more about what fixing this number actually means? AFAIK, we need to specify this when a serializable class is subject to evolutions and we want to ensure that serialized instances of an old version of a class can be reloaded by a newer version of that same class. If that constant isn't present, then the JVM will compute one automatically from the class' structure. So, if we fix these UIDs, that means that from there on we have to take great care that future versions of these classes are structure-compatible with older versions, i.e. no field renaming or deletion, and no new field whose value is set up in the constructor. So my feeling is that this could cause more harm than good (having a serialization exception is better than some other obscure failure) furthermore considering that Cocoon is not a system where serialized objects are supposed to be long-lived. And in the hypothetical case where that would be the case, then the relevant classes, and only these ones, could have a serialVersionUID, with the appropriate warning in the javadoc about the needed compatibility of future versions. So my advice: disable the warning in Eclipse, and that's all! So I did. I just was not aware of the consequences. Now that I know how it works in more details I also think that this is a bad idea. -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65
Re: Eclipse warning concerning serializable classes
On Saturday 20 November 2004 08:16, Vadim Gritsenko wrote: No! 1 is not The stream-unique identifier. Vadim, I don't know where you get that 1L wouldn't work... But it does (and I kind of know what I am talking about). The serialVersionUID is used to determine the compatibility between a loaded class and the serialized object of such class. There is not much more to it. The serialver executable is provided so that you can figure out the serialVerUID after the fact of releasing your code first time. So, i.e. you can use any number you want, and your challenge is to figure out when the number needs to be changed, i.e. serialization compatibility is no longer possible. To that effect, by not using writeObject()/readObject() it will indeed be slightly more tricky to maintain the serialization compatibility. With those methods, it shouldn't be too hard to maintain compatibitity almost forever. Cheers Niclas -- +--//---+ / http://www.bali.ac/ / http://niclas.hedhman.org / +--//---+
Re: Eclipse warning concerning serializable classes
Niclas Hedhman wrote: On Saturday 20 November 2004 08:16, Vadim Gritsenko wrote: No! 1 is not The stream-unique identifier. Vadim, I don't know where you get that 1L wouldn't work... But it does (and I kind of know what I am talking about). Only from the Sun doc - I've not taken my time reading their serialization source code. Doc gives impression that it is gotta be universally unique number. The serialVersionUID is used to determine the compatibility between a loaded class and the serialized object of such class. There is not much more to it. The serialver executable is provided so that you can figure out the serialVerUID after the fact of releasing your code first time. Yup. So, i.e. you can use any number you want, and your challenge is to figure out when the number needs to be changed, i.e. serialization compatibility is no longer possible. True. To that effect, by not using writeObject()/readObject() it will indeed be slightly more tricky to maintain the serialization compatibility. With those methods, it shouldn't be too hard to maintain compatibitity almost forever. Yup. Vadim
Re: Eclipse warning concerning serializable classes
Ugo Cei wrote: Then go for it! I hate those warnings and was planning to remove some of them myself, like unread local variables or declared exceptions that are never thrown. There is a caveat: if you remove declared but not thrown exception, you'll break all classes which use this method. Vadim
Re: Eclipse warning concerning serializable classes
Il giorno 20/nov/04, alle 18:45, Vadim Gritsenko ha scritto: Ugo Cei wrote: Then go for it! I hate those warnings and was planning to remove some of them myself, like unread local variables or declared exceptions that are never thrown. There is a caveat: if you remove declared but not thrown exception, you'll break all classes which use this method. Yes, that is a possibly incompatible change, if the method is called from outside Cocoon. One more reason to have a formal distinction between public and internal classes. A bigger problem is derived classes overriding methods from which throws clauses were removed. One more reason to avoid checked exceptions ;-) I'd like to go over these cases one by one and start by fixing some of them in trunk, if it looks safe to do so. Then, if someone complains, we can always revert the change. Ugo -- Ugo Cei - http://beblogging.com/ smime.p7s Description: S/MIME cryptographic signature
Re: Eclipse warning concerning serializable classes
On Sunday 21 November 2004 01:38, Vadim Gritsenko wrote: Only from the Sun doc - I've not taken my time reading their serialization source code. Doc gives impression that it is gotta be universally unique number. :o) Your excuse is accepted... To brief you why I happen to know; I used to be heavily involved in RMI/Serialization back in the late 90's. I had my own kind-of RMI implementation under JDK1.0, so when RMI was released n JDK1.1, I was quick to throw away my own garbage and use RMI extremely early, and got to know a fair bit of the inner workings early on, and strong participation in the RMI community. That resulted in being invited by Sun to participate in JSR-78 RMI Custom Remote Reference (http://www.jcp.org/en/jsr/detail?id=78). Anecdote; The spec was forced through the process too quickly to be ready for the JDK1.4, and had parts that didn't belong (thread pooling and configuration), and that was a shame. Otherwise a really cool thing, which would allow for JSR-76 RMI Security. Most of JSR78 ended up in the Jini community, but I haven't followed that... Cheers Niclas -- +--//---+ / http://www.bali.ac/ / http://niclas.hedhman.org / +--//---+
Re: Eclipse warning concerning serializable classes
Hi Lezsek wich version of eclipse are you using? Best Regards, Antonio Gallardo Leszek Gawron dijo: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; This is one of eclipse's QuickFix so I can have this done in few minutes. -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65
Re: Eclipse warning concerning serializable classes
Antonio Gallardo wrote: Hi Lezsek wich version of eclipse are you using? Best Regards, Antonio Gallardo Leszek Gawron dijo: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; This is one of eclipse's QuickFix so I can have this done in few minutes. -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65 3.1.0 build 200408122000 (win32) -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65
Re: Eclipse warning concerning serializable classes
Leszek Gawron wrote: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; No! 1 is not The stream-unique identifier. http://java.sun.com/j2se/1.4.2/docs/guide/serialization/spec/class.html Vadim
Re: Eclipse warning concerning serializable classes
Vadim Gritsenko wrote: Leszek Gawron wrote: The serializable class PageLocalImpl does not declare a static final serialVersionUID field of type long There is a lot of these in cocoon's sources. Can I fix that by adding: private static final long serialVersionUID = 1L; No! 1 is not The stream-unique identifier. http://java.sun.com/j2se/1.4.2/docs/guide/serialization/spec/class.html This was just an example :). Eclipse has an option to insert a generated serialVersionUID. Thing is after your post I checked it and it .. does not do anything :) -- Leszek Gawron [EMAIL PROTECTED] Project ManagerMobileBox sp. z o.o. +48 (61) 855 06 67 http://www.mobilebox.pl mobile: +48 (501) 720 812 fax: +48 (61) 853 29 65
