83544 - redundant global functions in smime & vcard

Status: VERIFIED FIXED

(MailNews Core :: MIME, defect)

Description

[cls]

In attempting to link all of the mail components into a meta component (bug
46775), I ran across a problem with the same functions being defined/declared
extern "C" in both vcard & smime. Similar to bug 76788 I think.

../../dist/lib/components/libsmime.a(mimexpcom.o): In function
`COM_GetmimeInlineTextClass':
mimexpcom.o(.text+0x0): multiple definition of `COM_GetmimeInlineTextClass'
../../dist/lib/components/libvcard.a(mimexpcom.o)(.text+0x0): first defined here


../../dist/lib/components/libsmime.a(nsSMIMEStub.o): In function
`MIME_GetContentType':
nsSMIMEStub.o(.text+0x0): multiple definition of `MIME_GetContentType'
../../dist/lib/components/libvcard.a(mimevcrd.o)(.text+0x0): first defined here

etc

Comment 1

[Chris Waterson]

This was fixed when we landed the STATIC_BUILD_20010612_BRANCH, no? (If not
before?) cls, re-open the bug if I'm wrong...

Comment 2

[cls]

This was never fixed on the branch.   All I did was reomve smime from the list
of components to be included in the mail meta-component.   I'm guessing the
reason that we're not seeing this problem in the completely static build is that
those functions in smime and/or vcard are never directly referenced so they are
stripped out in the final link.

Comment 3

[Chris Waterson]

Ok, sorry for jumping the gun.

Comment 4

[Suresh Duddi (gone)]

How is this a perf issue - symbol fixup when loading ?

Comment 5

[Chris Waterson]

I think this was blocking the static build at one point.

Comment 6

[Jean-Francois Ducarroz]

Every mime cthandlers (calendar, cvard, smime) include a file name
mimexpcom.cpp. They are all the same therefore, in the static build you need to
make sure you include only one of them!

Comment 7

[Jean-Francois Ducarroz]

Attachment: proposed fix, v1

Comment 8

[hacker formerly known as [e-mail]]

Comment on attachment 54922 [details] [diff] [review]
proposed fix, v1

It didn't quite work as the nsMimeContentTypeHandler.* files are duplicated as well.

Comment 9

[hacker formerly known as [e-mail]]

With cthandlers/*/mimexpcom.* & cthandlers/*/nsMimeContentTypeHandler.* being
exact duplicates, I decided to fork them off into a static glue lib that would
be linked into each of the handlers in a dynamic build but only into the final
link of a static build.  I ran into a slight problem because each of the 3
cthandlers uses the single nsMimeContentTypeHandler class as their registered
module.  So I had to hack the factories & nsMimeContentTypeHandler to allow the
factories to pass in mimetypes & a callback function to distinguish between the
cthandlers.  Surprisingly, it even worked.

Here's a patch against cthandlers/ , a separate patch of the changes made to
nsMimeContentTypeHandler and a zip of cthandlers/glue .

Comment 10

[hacker formerly known as [e-mail]]

Attachment: Modify cthandler factories to use modified nsMimeContentTypeHandler class

Comment 11

[hacker formerly known as [e-mail]]

Attachment: Changes to nsMimeContentHandlerType class (files are actually in glue/ now)

Comment 12

[hacker formerly known as [e-mail]]

Attachment: zip of glue/

Comment 13

[hacker formerly known as [e-mail]]

Attachment: Updated with win32 changes

Comment 14

[hacker formerly known as [e-mail]]

Attachment: Updated glue.zip with mac+win32 changes

Comment 15

[hacker formerly known as [e-mail]]

Attachment: calendar.mcp

Comment 16

[hacker formerly known as [e-mail]]

Attachment: smime.mcp

Comment 17

[hacker formerly known as [e-mail]]

Attachment: vcard.mcp

Comment 18

[Jean-Francois Ducarroz]

Christopher, can I reassign this bug to you as you do all the work?

Comment 19

[hacker formerly known as [e-mail]]

Sure, I'll take it.  But I still need a review / verification of the patches
since this isn't exactly my area of expertise. :)

Comment 20

[Jean-Francois Ducarroz]

Comment on attachment 56288 [details] [diff] [review]
Updated with win32 changes

R=ducarroz

Comment 21

[Jean-Francois Ducarroz]

Comment on attachment 55214 [details] [diff] [review]
Changes to nsMimeContentHandlerType class (files are actually in glue/ now)

R=ducarroz

Comment 22

[Jean-Francois Ducarroz]

Comment on attachment 56292 [details]
vcard.mcp

R=ducarroz

Comment 23

[Jean-Francois Ducarroz]

Comment on attachment 56291 [details]
smime.mcp

R=ducarroz

Comment 24

[Jean-Francois Ducarroz]

Comment on attachment 56290 [details]
calendar.mcp

R=ducarroz

Comment 25

[Jean-Francois Ducarroz]

Comment on attachment 56289 [details]
Updated glue.zip with mac+win32 changes

R=ducarroz

Comment 26

[Chris Waterson]

rs=waterson

Comment 27

[hacker formerly known as [e-mail]]

Checked in.

Comment 28

[[e-mail] (gone - use [e-mail] instead)]

By verifying this, I'm assuming that if you can build successfully (which I
have) then it's just something to verify via code inspection (using LXR/tree).

If I'm wrong, please re-open and correct me.