Re: [sqlite] endianess/signed issue on OpenBSD/sparc64 ?
On Thu, Jun 13 2019, Richard Hipp wrote: > On 6/9/19, Landry Breuil wrote: >> >> this broke on sparc64 >>>select 298.2564151; >> -298.2564151 >> > > I cannot reproduce this (perhaps because I do not have access to a > sparc64 platform running OpenBSD) and do not have any good ideas about > what might be causing it. If you can get me a temporary login on such > a system, however, I will debug it for you. Send private email to the > address below. This was indeed not a bug in SQLite, but in a sparc64-specific part of the OpenBSD libc, as pointed out privately by Richard. I have committed a fix (picked up from FreeBSD): https://marc.info/?l=openbsd-cvs=156113647005489=2 Richard, thanks a lot for your time and expertise. -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE signature.asc Description: PGP signature ___ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
Re: [sqlite] endianess/signed issue on OpenBSD/sparc64 ?
On Thu, Jun 13 2019, Richard Hipp wrote: > On 6/9/19, Landry Breuil wrote: >> >> this broke on sparc64 >>>select 298.2564151; >> -298.2564151 >> > > I cannot reproduce this (perhaps because I do not have access to a > sparc64 platform running OpenBSD) and do not have any good ideas about > what might be causing it. If you can get me a temporary login on such > a system, however, I will debug it for you. Send private email to the > address below. Thanks for the proposal, I have sent you a private mail. For the record: - I have built a bunch of older sqlite3 releases: 3.8.2, 3.16.2, 3.24.0, 3.25.1, 3.25.3 and 3.27.2. All of them give the same wrong result for "select 298.2564151;" - on OpenBSD/sparc64 both the base system and sqlite3 are built using gcc-4.2.1, an old gcc release licensed under GPLv2. Using gcc-8.3.0 to build sqlite3 doesn't magically fix the problem. -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE signature.asc Description: PGP signature ___ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
Re: [sqlite] endianess/signed issue on OpenBSD/sparc64 ?
On 6/9/19, Landry Breuil wrote: > > this broke on sparc64 >>select 298.2564151; > -298.2564151 > I cannot reproduce this (perhaps because I do not have access to a sparc64 platform running OpenBSD) and do not have any good ideas about what might be causing it. If you can get me a temporary login on such a system, however, I will debug it for you. Send private email to the address below. -- D. Richard Hipp d...@sqlite.org ___ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
[sqlite] endianess/signed issue on OpenBSD/sparc64 ?
Hi (not subscribed to the list), i'm more or less maintaining sqlite3 in OpenBSD packages (at 3.28.0 now), and recently proj.4 was updated to a version that relies on sqlite3 to store projections, which has some constraints on values entered. this broke on sparc64 (cf https://github.com/OSGeo/PROJ/issues/1508), because for some reason a float value gets converted to its opposite, which show simply with (per https://github.com/OSGeo/PROJ/issues/1508#issuecomment-500141325): >select 298.2564151; -298.2564151 of course, on OpenBSD/amd64 this simple sql statement works. So a bug in sqlite itself ? I don't have access to a sparc64 system right now, but i'd try previous sqlite versions to check for a regression. Landry ___ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users