Thread: ecpg leaves broken files around
When working on the ecpg regression stuff things, it's the first time I've actually used ecpg. And I notice now that it leaves incomplete and broken files around when it fails. For example, I currently get: parser.pgc:26: ERROR: syntax error at or near "NULLS" when building. I have to investigate why later, but when I run ecpg on this it leaves a parser.c file around, which the build system on the next run picks up as there and doesn't try to rebuild. Naturally, it fails later when the C compiler tries to build this file. AFAIK, most other compilers delete their output if it's not valid. Is there any particular reason why ecpg doesn't do this? //Magnus
On Sun, Jun 10, 2007 at 09:56:44PM +0200, Magnus Hagander wrote: > AFAIK, most other compilers delete their output if it's not valid. Is > there any particular reason why ecpg doesn't do this? No, not really. Sometimes it comes handy to see what was already processed, but you're right, it's not what I would expect from a compiler either. Any objects changing this behaviour? Michael -- Michael Meskes Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org) ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!
On Mon, Jun 11, 2007 at 08:05:16AM +0200, Michael Meskes wrote: > On Sun, Jun 10, 2007 at 09:56:44PM +0200, Magnus Hagander wrote: > > AFAIK, most other compilers delete their output if it's not valid. Is > > there any particular reason why ecpg doesn't do this? > > No, not really. Sometimes it comes handy to see what was already > processed, but you're right, it's not what I would expect from a > compiler either. > > Any objects changing this behaviour? Certainly not from me :) If you find the other behaviour useful, perhaps add a commandline switch that makes it leave the file around? Just make the remove-the-file-on-failure default. Oh, and it seems if you want to keep the feature, it needs fixing. It looks like ecpg doesn't flush/close the file descriptor before error-exit, so the file that drops out isn't even complete up to the point of error. At least that's what it looks like from a quick glance - it ends mid-row on something that's not related to the error itself. //Magnus
On Mon, Jun 11, 2007 at 10:03:57AM +0200, Magnus Hagander wrote: > If you find the other behaviour useful, perhaps add a commandline switch > that makes it leave the file around? Just make the > remove-the-file-on-failure default. Should be fixed now. I don't think such a command line switch is needed. After all you could run ecpg with "-o -" and make it output to stdout. Michael -- Michael Meskes Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org) ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!