Обсуждение: nightly win32 builds
I've been trying to add --with-openssl options to the nightly builds, but have not been able to get postgresql compiled with this option. MinGW compiles the openssl library with no problems... configure --with-openssl configure complains about missing crypto library ok, I manually added --with-libs=/usr/local/ssl/lib (even though this is the default location), now crypto is found, but configure still complains about missing ssl library, which is in the same folder. (this is all with openssl-0.9.7d, btw). I was curious, so I manually hacked configure to skip the ssl library check so I could see what happened during the build. Never made it to the link step, there were problems trying to compile the front end. I gave up at that point, thinking maybe I picked a openssl library that was too new. Has anybody had any luck compiling recent version of pg with openssl (either in mingw or otherwise?) As a side note, I bumped up gcc to 3.4 trying to resolve this. It didn't help, but did result in a 2/3 reduction in compile times, so I'm leaving it that way until somebody says otherwise :) Merlin
I would look in config.log to see why configure is failing. --------------------------------------------------------------------------- Merlin Moncure wrote: > I've been trying to add --with-openssl options to the nightly builds, > but have not been able to get postgresql compiled with this option. > MinGW compiles the openssl library with no problems... > > configure --with-openssl > configure complains about missing crypto library > > ok, I manually added --with-libs=/usr/local/ssl/lib (even though this is > the default location), now crypto is found, but configure still > complains about missing ssl library, which is in the same folder. (this > is all with openssl-0.9.7d, btw). > > I was curious, so I manually hacked configure to skip the ssl library > check so I could see what happened during the build. Never made it to > the link step, there were problems trying to compile the front end. I > gave up at that point, thinking maybe I picked a openssl library that > was too new. > > Has anybody had any luck compiling recent version of pg with openssl > (either in mingw or otherwise?) > > As a side note, I bumped up gcc to 3.4 trying to resolve this. It > didn't help, but did result in a 2/3 reduction in compile times, so I'm > leaving it that way until somebody says otherwise :) > > Merlin > > > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
> I would look in config.log to see why configure is failing. The problem is that openssl is making some calls to win32 GDI code in a function called read_screen, which allows the user to generate a random number based in UI interaction. These are not linked in properly...this is a mingw/openssl problem. It is very unlikely this function is used by postgres, so I commented the code and configure passes muster (except that pg assumes openssl for the root ssl install folder, and openssl seems to default to ssl). The upshot is you have to manually add a couple of folders to the configure statement, no big deal. There are a couple of other problems with ssl support in be-secure.c which I have fixed. There were more problems in exec.c... validate_exec() should be declared with const char*, not char char* (for gcc 3.4) and there were problems using the palloc memory context macros (can't import CurrentMemoryContext). I hacked around this by using the stack, but this error popped up in several more places so I gave up... Merlin
Ok, I got ssl built, finally. I had to fix a minor problem in exec.c and everything seems to work ok. If anybody is interested the binary is available here: http://www.hagander.net/pgsql/win32snap/ Merlin
Merlin, thanks for making that available. Tyler pgsql-hackers-win32-owner@postgresql.org wrote on 05/21/2004 06:29:39 AM: > > Ok, I got ssl built, finally. I had to fix a minor problem in exec.c > and everything seems to work ok. If anybody is interested the binary is > available here: > > http://www.hagander.net/pgsql/win32snap/ > > Merlin > > ---------------------------(end of broadcast)--------------------------- > TIP 7: don't forget to increase your free space map settings
Merlin Moncure wrote: > > Ok, I got ssl built, finally. I had to fix a minor problem in exec.c > and everything seems to work ok. If anybody is interested the binary is > available here: > > http://www.hagander.net/pgsql/win32snap/ I would like to have patches against current CVS so these fixes are in 7.5. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
> > Ok, I got ssl built, finally. I had to fix a minor problem in exec.c > > and everything seems to work ok. If anybody is interested the binary is > > available here: > > > > http://www.hagander.net/pgsql/win32snap/ > > I would like to have patches against current CVS so these fixes are in > 7.5. According to Magnus, there is a good chance that signals coming in while the backend is waiting on a ssl connection will have similar issues that had to be resolved for the main waiting function. I need to confirm this and (with Magnus's help) perhaps work this into the patch. Merlin
Merlin Moncure wrote: > > > Ok, I got ssl built, finally. I had to fix a minor problem in > exec.c > > > and everything seems to work ok. If anybody is interested the > binary is > > > available here: > > > > > > http://www.hagander.net/pgsql/win32snap/ > > > > I would like to have patches against current CVS so these fixes are in > > 7.5. > > According to Magnus, there is a good chance that signals coming in while > the backend is waiting on a ssl connection will have similar issues that > had to be resolved for the main waiting function. I need to confirm > this and (with Magnus's help) perhaps work this into the patch. OK, just send what you have when you are read. I think I might have fixed a few of the problems you already reported so be sure to try the CVS for that day before sending the patch. Win32 is getting lots of activity. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073