Fix output messages

[lucmaga]

• Add missing newlines
• Use errlogPrintf for errors

fixes #81

[simon-ess]

There seem to be a few also in the function save_restoreShow, see

            epicsTimeToStrftime(tmpstr, sizeof(tmpstr), TIMEFMT, &plist->save_time);
            printf("    Last save time  :%s", tmpstr);
            epicsTimeToStrftime(tmpstr, sizeof(tmpstr), TIMEFMT, &plist->backup_time);
printf(" Last backup time:%s", tmpstr);

and

printf(" channel_connected = %d", pchannel->channel_connected);

[anjohnson]

Please don't mindlessly convert every printf() call into errlogPrintf() without looking at them carefully.

I'm not sure whether you have or not since there are so many changes here and I don't know the code in detail, but any routines that are normally run by an IOCSH command (possibly indirectly) may be deliberately using printf() so those messages are displayed synchronously with the command execution. Printing shell-command error messages through errlog can cause them to be interleaved with other command outputs and get out of order, which gets confusing to the user. I'm specifically looking at the "manual" routines (there are many IOCSH commands with that in the name), although I don't know whether they can be triggered from a PV or are just called from IOCSH commands. That might apply to other routines too though.

Please respond and confirm whether you were aware of and have considered this subtlety.

Also, the filename "log.h" makes me a bit nervous given how generic that is, although I believe it should always be found before any OS-specific "log.h" file since you're using double-quotes in the #include statement for it.

[lucmaga]

Please don't mindlessly convert every printf() call into errlogPrintf() without looking at them carefully.

I'm not sure whether you have or not since there are so many changes here and I don't know the code in detail, but any routines that are normally run by an IOCSH command (possibly indirectly) may be deliberately using printf() so those messages are displayed synchronously with the command execution. Printing shell-command error messages through errlog can cause them to be interleaved with other command outputs and get out of order, which gets confusing to the user. I'm specifically looking at the "manual" routines (there are many IOCSH commands with that in the name), although I don't know whether they can be triggered from a PV or are just called from IOCSH commands. That might apply to other routines too though.

Please respond and confirm whether you were aware of and have considered this subtlety.

Thanks for the review. The commits 59d6185 and 59736ee are the ones I convert printf to errlogPrintf. I kept them separate for easier review.
I was careful to convert only messages where there was a branch checking on error, a return ERROR like statement. Specially save_restore.c is a tricky one to convert, I'm not sure I cover all error outputs but most of them, at least.

Also, the filename "log.h" makes me a bit nervous given how generic that is, although I believe it should always be found before any OS-specific "log.h" file since you're using double-quotes in the #include statement for it.

Funny, I was expecting this. I'll try to think into a better name.

[simon-ess]

I think that this is overall a good change, although as @anjohnson pointed out there are a few of the errlog changes that perhaps should remain printfs (in the manual restore function).

[simon-ess]

LGTM now. Maybe add a release note about this, noting that anyone parsing their IOC logs (e.g. ESS 😄 ) for saverestore output may want to revisit that parsing...

[lucmaga]

Hi, is there anything missing for this to be merged?