[PATCH] user_ns: compile bug fix in quota.h

[PATCH] user_ns: compile bug fix in quota.h

[Zhao Hongjiang]

From: Zhao Hongjiang <zhaohongjiang-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>

Failed to build userns-always-map-user-v53 from Eric's latest
user_ns branch with GCC 4.4.4 compiler due to violation of
C89 standard.

Signed-off-by: Zhao Hongjiang <zhaohongjiang-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>
---
 include/linux/quota.h |   15 +++++++++------
 1 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/include/linux/quota.h b/include/linux/quota.h
index 781ffc4..90f12d7 100644
--- a/include/linux/quota.h
+++ b/include/linux/quota.h
@@ -280,9 +280,10 @@ static inline struct kqid make_kqid_invalid(enum quota_type type)
 static inline struct kqid make_kqid_uid(kuid_t uid)
 {
 	struct kqid kqid = {
-		.type = USRQUOTA,
-		.uid  = uid,
+		.type = USRQUOTA
 	};
+	kqid.uid = uid;
+
 	return kqid;
 }

@@ -293,9 +294,10 @@ static inline struct kqid make_kqid_uid(kuid_t uid)
 static inline struct kqid make_kqid_gid(kgid_t gid)
 {
 	struct kqid kqid = {
-		.type = GRPQUOTA,
-		.gid  = gid,
+		.type = GRPQUOTA
 	};
+	kqid.gid  = gid;
+
 	return kqid;
 }

@@ -306,9 +308,10 @@ static inline struct kqid make_kqid_gid(kgid_t gid)
 static inline struct kqid make_kqid_projid(kprojid_t projid)
 {
 	struct kqid kqid = {
-		.type = PRJQUOTA,
-		.projid = projid,
+		.type = PRJQUOTA
 	};
+	kqid.projid = projid;
+
 	return kqid;
 }

-- 1.7.1

Re: [PATCH] user_ns: compile bug fix in quota.h

[Eric W. Biederman]

Zhao Hongjiang <zhaohongjiang37-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> writes:

> From: Zhao Hongjiang <zhaohongjiang-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>
>
> Failed to build userns-always-map-user-v53 from Eric's latest
> user_ns branch with GCC 4.4.4 compiler due to violation of
> C89 standard.

Thanks.

Given that they are C99 style initializers in general I don't see how
taking one member out of the initializer and leaving the other solves
the C89 violation.  But I do remember that gcc has historically had some
interesting problems with C99 initializers, so in this case I have just
opted to not use an initializer for either field, which should
make older gcc versions.

Eric