550474 - Fix '--enable-static --disable-static-mail' behavior

Status: RESOLVED FIXED

(MailNews Core :: Build Config, defect)

Description

[Serge Gautherie (:sgautherie)]

Options are:

--enable-static sets BUILD_STATIC_LIBS=1.
default = disabled in configure.in.

--disable-static-mail sets MOZ_STATIC_MAIL_BUILD=0.
default = enabled in SM+TB confvars.sh.

***

Relevant code examples:

configure.in:
    if test -n "$BUILD_STATIC_LIBS"; then
        AC_DEFINE(MOZ_STATIC_BUILD)
    fi

http://mxr.mozilla.org/comm-central/source/mailnews/base/src/nsMessengerWinIntegration.cpp#312
313 #ifdef MOZ_STATIC_BUILD
314 #define MAIL_DLL_NAME NULL
315 #else
316 #ifdef MOZ_STATIC_MAIL_BUILD
317 #define MAIL_DLL_NAME "mail.dll"
318 #else
319 #define MAIL_DLL_NAME "msgbase.dll"
320 #endif
321 #endif

*****

Build results:

-static + -static-mail:
1) Y + Y = all dlls merged into <application>.exe
2) N + Y = individual dlls, but with all mailnews dlls merged into components/mail.dll
3) Y + N = (!? see below)
4) N + N = individual dlls, mailnews ones too

***

Do we actually need the *-static-mail option (in addition to the *-static one)?
(In which case is it useful, for example?)

If we keep this option, what should '--enable-static --disable-static-mail' result be?
*Not supported: add an error check in configure.in.
*Supported: fix code so it builds correctly.

Comment 1

[Serge Gautherie (:sgautherie)]

At the moment, trying to build SeaMonkey 2.0.4pre with these options fail at:
{
   Creating library TestImapHdrXferInfo.lib and object TestImapHdrXferInfo.exp
msgimap_s.lib(nsImapProtocol.obj) : error LNK2001: unresolved external symbol "public: virtual unsigned int __stdcall nsRDFResource::QueryInterface(struct nsID const &,void * *)" (?QueryInterface@nsRDFResource@@UAGIABUnsID@@PAPAX@Z)
[...]
}

Comment 2

[Robert Kaiser]

IMHO, --enable-static --disable-static-mail should error out in configure.
I'll let Mark or Dan make the final call on this, though.

Comment 3

[Mark Banner (:standard8)]

(In reply to comment #0)
> Do we actually need the *-static-mail option (in addition to the *-static one)?
> (In which case is it useful, for example?)

Yes. It means that we can run our unit test and bloat builders in a non-static configuration (which bloat & unit test don't currently support) *and* in that configuration we can run mailnews as close as possible to the static configuration - i.e. using mailnews/build for the factory registration code which is what is used in static build configuration (this is versus the individual mailnews/<foo>/build directories used in the completely non-static configuration).

And yes we do forget to sync stuff between those at times, so having our tinderboxes run the stuff we're going to ship is the better option.

We also don't have static mail by default all the time because various developers complain about the time to recompile etc.

> If we keep this option, what should '--enable-static --disable-static-mail'
> result be?
> *Not supported: add an error check in configure.in.
> *Supported: fix code so it builds correctly.

This isn't something I care much about as I don't see devs likely to do it too often, but I guess we should just error out in configure.in.

Comment 4

[Serge Gautherie (:sgautherie)]

Attachment: (Av1) Explicitly error out [Checkin: See comment 6]

Comment 5

[Justin Wood (:Callek)]

Comment on attachment 430941 [details] [diff] [review]
(Av1) Explicitly error out
[Checkin: See comment 6]

per bug comments on desired outcome, looks good.

Comment 6

[Serge Gautherie (:sgautherie)]

Comment on attachment 430941 [details] [diff] [review]
(Av1) Explicitly error out
[Checkin: See comment 6]


http://hg.mozilla.org/comm-central/rev/98495510718d
Av1, with s/tab/spaces/.


"approval-thunderbird3.0.4=?":
Zero risk, build config only.

Comment 7

[Mark Banner (:standard8)]

Comment on attachment 430941 [details] [diff] [review]
(Av1) Explicitly error out
[Checkin: See comment 6]

I don't care about this for 3.0.x, I think that the amount of people likely to hit this versus the addition bits of work to land and verify don't warrant it.