Re: CVS sources doesn't compiles

On Tue, 2002-08-06 at 14:49, Tom Lane wrote:
> John Gray <jgray(at)azuli(dot)co(dot)uk> writes:
> > On Tue, 2002-08-06 at 14:03, Oleg Bartunov wrote:
> >> make[4]: Entering directory /db1/pgsql/cvs/pgsql-server/src/backend/access/heap'
> >> gcc -O2 -mpentiumpro -Wall -Wmissing-prototypes -Wmissing-declarations -I../../../../src/include -c -o heapam.o heapam.c
> >> heapam.c: In function eap_insert':
> >> heapam.c:1158: structure has no member named d_istemp'
> >> heapam.c: In function eap_delete':
> >> heapam.c:1341: structure has no member named d_istemp'
> >> heapam.c: In function eap_update':
> >> heapam.c:1677: structure has no member named d_istemp'
> >> make[4]: *** [heapam.o] Error 1
> >> make[4]: Leaving directory /db1/pgsql/cvs/pgsql-server/src/backend/access/heap'
>
> Control-H? Control-R? You seem to have a corrupted copy of heapam.c.
> If you move it out of the way and do a "cvs update", do you get a copy
> with the identical errors?
>

I should have checked what I was quoting first! The messages I get have
no funny characters in -and the reason for the error is that rel.h has
the following (as you can see, it doesn't have rd_istemp replacing
rd_myxactonly):

/*
* Here are the contents of a relation cache entry.
*/

typedef struct RelationData
{
File rd_fd; /* open file descriptor,
or -1 if none */
RelFileNode rd_node; /* file node (physical
identifier) */
BlockNumber rd_nblocks; /* number of blocks in rel */
BlockNumber rd_targblock; /* current insertion target
block, or
*
InvalidBlockNumber */
int rd_refcnt; /* reference
count */
bool rd_myxactonly; /* rel uses the local buffer mgr
*/
bool rd_isnailed; /* rel is nailed in cache */
bool rd_indexfound; /* true if rd_indexlist is valid
*/
bool rd_uniqueindex; /* true if rel is a UNIQUE index
*/

[rest snipped]. This is version 1.60. Tom's patch produced 1.61. I can't
get anonCVS to give me 1.61. (But annoyingly, it gives me Tom's updated
heapam.c 1.144).

> I can report that the master CVS server delivers a correct copy. If
> there is a CVS problem then it's only on the anoncvs mirror ...
>

Well, that seems likely -as cvsweb reports the file OK.

I wonder whether our CVS mirroring is sufficiently atomic? i.e. did we
get an inconsistent snapshot because it was taken partway through a
patch being applied.

This is clearly going to be a bit of a pain if it is a consequence of
heavier development activity - not least because it consumes everyone's
time chasing imaginary bugs. I'm assuming that it will just be a
transient issue - but there has been no change in it for several hours,
so presumably the mirroring is not run that often...

Regards

John

--
John Gray
Azuli IT
www.azuli.co.uk