From: Tom Lane <tgl@sss.pgh.pa.us>
To: Thomas Munro <thomas.munro@gmail.com>
cc: Peter Eisentraut <peter@eisentraut.org>, Bruce Momjian <bruce@momjian.us>,
        walther@technowledgy.de, Christophe Pettus <xof@thebuild.com>,
        Andrew Dunstan <andrew@dunslane.net>,
        PostgreSQL Bugs <pgsql-bugs@lists.postgresql.org>,
        Robert Haas <robertmhaas@gmail.com>
Subject: Re: Regression tests fail with musl libc because libpq.so can't be
 loaded
In-reply-to: <2712381.1711408552@sss.pgh.pa.us>
References: 
 <CA+hUKGK2cSx3dveMqBD+E9N=zJ0WawBcbk_jnPqmdWBM0DT_3Q@mail.gmail.com>
 <477fa513-6bf0-456d-8d24-1e22a19df671@technowledgy.de>
 <ZfrmN4WHuJgD95fw@momjian.us>
 <CA+hUKGJzC4rr2Yr=QZyx9W83Wv=TVJHndKF1anjNLNXVUM7TqQ@mail.gmail.com>
 <f98cd8de-1c66-491a-8409-e62c09932080@technowledgy.de>
 <36690fcf-f0c8-4d60-af49-6a7efa0e2b14@technowledgy.de>
 <CA+hUKGJAkTZUa9JGr-zsadaqsN=K0P71t-1L2t2=MQFDMc4GuQ@mail.gmail.com>
 <ZfzLsiXbguTDatiZ@momjian.us>
 <a2463df1-be45-4a76-b90f-7676d778e7b1@technowledgy.de>
 <Zf2JU0_GDdbeBm5j@momjian.us> <Zf3fnKpHIGpe5mOi@momjian.us>
 <443430b4-0c10-49f3-8c8a-e5e76602f309@eisentraut.org>
 <CA+hUKGKU_0hmqem_EmAwePjVtMrjZNyNN0niUdBhuABajRwyhw@mail.gmail.com>
 <2712381.1711408552@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Mon, 25 Mar 2024 19:15:52 -0400"
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="----- =_aaaaaaaaaa0"
Content-ID: <2715709.1711410168.0@sss.pgh.pa.us>
Date: Mon, 25 Mar 2024 19:43:08 -0400
Message-ID: <2715727.1711410188@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/2715727.1711410188%40sss.pgh.pa.us>
Precedence: bulk

------- =_aaaaaaaaaa0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2715709.1711410168.1@sss.pgh.pa.us>

I wrote:
> Thomas Munro <thomas.munro@gmail.com> writes:
>> I had originally proposed to avoid anything beginning "LD_" but Tom
>> suggested being more specific.  I doubt LD_PRELOAD can really hurt you
>> though (the linker probably only needs the value at the start by
>> definition, not at later dlopen() time (?)).

> Oh, good point.  So we could simplify the patch by only looking for
> LD_LIBRARY_PATH.

I looked at the musl source code you identified and confirmed that
only the LD_LIBRARY_PATH string is remembered in a static variable;
LD_PRELOAD is only accessed locally in that initialization function.
So we only need to do the attached.  (I failed to resist the
temptation to rewrite the comments.)

			regards, tom lane


------- =_aaaaaaaaaa0
Content-Type: text/x-diff;
	name="v2-0001-Don-t-clobber-LD_-environment-variables.patch";
	charset="us-ascii"
Content-ID: <2715709.1711410168.2@sss.pgh.pa.us>
Content-Description: v2-0001-Don-t-clobber-LD_-environment-variables.patch
Content-Transfer-Encoding: quoted-printable

diff --git a/src/backend/utils/misc/ps_status.c b/src/backend/utils/misc/p=
s_status.c
index 5d829e6e48..92cd2c7899 100644
--- a/src/backend/utils/misc/ps_status.c
+++ b/src/backend/utils/misc/ps_status.c
@@ -151,7 +151,33 @@ save_ps_display_args(int argc, char **argv)
 		for (i =3D 0; environ[i] !=3D NULL; i++)
 		{
 			if (end_of_area + 1 =3D=3D environ[i])
-				end_of_area =3D environ[i] + strlen(environ[i]);
+			{
+				/*
+				 * The musl runtime linker keeps a static pointer to the
+				 * initial value of LD_LIBRARY_PATH, if that is defined in the
+				 * process's environment. Therefore, we must not overwrite the
+				 * value of that setting and thus cannot advance end_of_area
+				 * beyond it.  Musl does not define any identifying compiler
+				 * symbol, so we have to do this for any Linux libc we don't
+				 * know is safe.
+				 */
+#if defined(__linux__) && (!defined(__GLIBC__) && !defined(__UCLIBC__ ))
+				if (strncmp(environ[i], "LD_LIBRARY_PATH=3D", 16) =3D=3D 0)
+				{
+					/*
+					 * We can overwrite the name, but stop at the equals sign.
+					 * Future loop iterations will not find any more
+					 * contiguous space, but we don't break early because we
+					 * need to count the total number of environ[] entries.
+					 */
+					end_of_area =3D environ[i] + 15;
+				}
+				else
+#endif
+				{
+					end_of_area =3D environ[i] + strlen(environ[i]);
+				}
+			}
 		}
 =

 		ps_buffer =3D argv[0];

------- =_aaaaaaaaaa0--