Horton

Thread

Commits

  1. Fix ecpg's mishandling of B'...' and X'...' literals.

  1. errors when there is a bit literal in ecpg

    Wang, Shenhao <wangsh.fnst@cn.fujitsu.com> — 2020-11-06T22:56:11Z

    Hi, hacker

I met an error as below when I use ecpg

a.pgc:5: ERROR: invalid bit string literal
a.pgc:5: ERROR: internal error: unreachable state; please report this to <pgsql-bugs@lists.postgresql.org>

the test source is attached.

After investigating the code, I think the process of pgc.l is:
Step 1:  <SQL>{xbstart},  addlitchar('b') is called, literalbuf contains a char 'b';
Step 2:  <xb>{xbinside},   the rest of char is added in literalbuf
Step 3:   <xqs>{other},    the    condition literalbuf[strspn(literalbuf, "01") + 1] != '\0'   will always be true;
error is occurred here

I try to fix this bug by deleting 'addlitchar('b');' from source I also add a test case to test all const str in ecpg.

The patch is also attached.

Best regards, Shenhao Wang

  2. Re: errors when there is a bit literal in ecpg

    Tom Lane <tgl@sss.pgh.pa.us> — 2020-11-07T20:05:42Z

    "Wang, Shenhao" <wangsh.fnst@cn.fujitsu.com> writes:
> I met an error as below when I use ecpg

> a.pgc:5: ERROR: invalid bit string literal
> a.pgc:5: ERROR: internal error: unreachable state; please report this to <pgsql-bugs@lists.postgresql.org>

Indeed.  This has apparently been broken for a very long time (though
the "unreachable state" part is fairly new).

> I try to fix this bug by deleting 'addlitchar('b');' from source I also add a test case to test all const str in ecpg.

I thought that a whole new test case was overkill when we could just add a
couple of lines to an existing test.  Other than that, looks good, pushed.

			regards, tom lane