beta2 make check failed on Win32

pgsql-8.1beta2 "make check" failed on Win32 under MinGW.
Here is some lines from log/initdb.log.

<skipped>
creating configuration files ... ok
creating template1 database in
C:/MSYS/home/wlzhang/postgresql-8.1beta2/src/test/regress/tmp_check/data/bas
e/1 ... FATAL: syntax error in file
"C:/MSYS/home/wlzhang/postgresql-8.1beta2/src/test/regress/tmp_check/data/po
stgresql.conf" line 386, near token "'s Republic of China.936'"
child process was terminated by signal 1
initdb: data directory
"C:/MSYS/home/wlzhang/postgresql-8.1beta2/src/test/regress/tmp_check/data"
not removed at user's request

And some lines from postgresql.conf.

# These settings are initialized by initdb -- they might be changed
lc_messages = 'Chinese_People''s Republic of China.936' # locale for
system error message

# strings
lc_monetary = 'Chinese_People''s Republic of China.936' # locale for
monetary formatting
lc_numeric = 'Chinese_People''s Republic of China.936' # locale for
number formatting
lc_time = 'Chinese_People''s Republic of China.936' # locale for
time formatting

It seems that we cannot deal with embedded quotes correctly.
For simplicity, we can test using any parameter with string type. e.g.
bonjour_name.

--
Regards,
William ZHANG

"William ZHANG" <uniware(at)zedware(dot)org> wrote

> lc_messages = 'Chinese_People''s Republic of China.936' # locale
> for
> system error message
>

Diff current version vs. some earlier version, you see this:

/*
* signal handler in case we are interrupted.
@@ -1951,12 +2035,12 @@ escape_quotes(const char *src)
{
int len = strlen(src),
i, j;
- char *result = xmalloc(len * 2 + 1);
+ char *result = pg_malloc(len * 2 + 1);

for (i = 0, j = 0; i < len; i++)
{
- if (src[i] == '\'' || src[i] == '\\')
- result[j++] = '\\';
+ if (SQL_STR_DOUBLE(src[i]))
+ result[j++] = src[i];
result[j++] = src[i];
}

So the problem is this line:

result[j++] = src[i];

That is, use the old line then the 'People\'s republic of China' string gets
right. But I am quite sure if revoke this will affect other things.

Regards,
Qingqing

What you find is in initdb.c. Maybe the author want to make single quotes
escaped just the same as in SQL. But guc-file.l is unaware of it.

"Qingqing Zhou" <zhouqq(at)cs(dot)toronto(dot)edu> wrote
>
> Diff current version vs. some earlier version, you see this:
>
> /*
> * signal handler in case we are interrupted.
> @@ -1951,12 +2035,12 @@ escape_quotes(const char *src)
> {
> int len = strlen(src),
> i, j;
> - char *result = xmalloc(len * 2 + 1);
> + char *result = pg_malloc(len * 2 + 1);
>
> for (i = 0, j = 0; i < len; i++)
> {
> - if (src[i] == '\'' || src[i] == '\\')
> - result[j++] = '\\';
> + if (SQL_STR_DOUBLE(src[i]))
> + result[j++] = src[i];
> result[j++] = src[i];
> }
>
> So the problem is this line:
>
> result[j++] = src[i];
>
> That is, use the old line then the 'People\'s republic of China' string
gets
> right. But I am quite sure if revoke this will affect other things.
>
> Regards,
> Qingqing
>
>
>
>

"William ZHANG" <uniware(at)zedware(dot)org> writes:
> # These settings are initialized by initdb -- they might be changed
> lc_messages = 'Chinese_People''s Republic of China.936' # locale for
> system error message

Bruce seems to have gotten a bit ahead of himself on the
SQL-standard-strings project --- guc-file.l did not actually accept
the above as legal syntax. Seems it should though; fixed accordingly.

regards, tom lane