Re: [Monotone-devel] Monotone 0.43dev build error, paths.hh missing

[Zack Weinberg]

On Fri, Mar 6, 2009 at 9:58 PM, Tero Koskinen <address@hidden> wrote:
> On Fri, 6 Mar 2009 20:19:39 -0700 Derek Scherger wrote:
>> On Fri, Mar 6, 2009 at 1:15 PM, Tero Koskinen <address@hidden> wrote:
>> > It seems that merge_content.hh is missing #include "paths.hh".
>>
>>
>> This is quite likely because I moved content_merge_workspace_adaptor to
>> merge_content.{cc,hh} from work.{cc,hh} in
>> dbf38fd6d5265df467a88a468d110bcc039c2367. Just to confirm, can you try
>> building add275e5fdb0824361d4a2fa15049281e7ada52e please?
>
> Building add275... failed also in the same place, complete log is at
> http://iki.fi/tero.koskinen/monotone/monotone-build-radd275e5fdb082.txt

content_merge_workspace_adaptor is still in merge_content.hh as of
revision add275... what you actually appear to have moved in dbf... is
content_merge_empty_adaptor.

I think it was actually me who moved content_merge_workspace_adaptor,
'way back in
d2d539af06eb22be04aabb884b41704320311761.

>> I'm not sure why I don't see the problem here though, gcc 4.1.2 seems
>> perfectly happy on a fresh checkout?
>
> I am using gcc 3.3.5 on OpenBSD. It might have some bugs which are
> fixed in gcc 4 series. (gcc 3.3.5 is the default system compiler,
> I use it because all C and C++ libraries are compiled with it.)

I've seen this problem before.  Some versions of g++/libstdc++ will
not allow you to declare a variable of type map<K,V> unless both K and
V are complete types at the point of declaration.  I don't know if
this is a bug or not.  In any case, working around it is no real
trouble.  I've committed Tero's patch.

Which version of OpenBSD is this, though?  gcc 3.3 predates the C++
parser rewrite that made it actually, er, parse C++ correctly, and I
would kinda like to drop support for it (there are a few places where
we have special hacks just for gcc <=3.3).

zw