How to log the filename of the file printed by cups?

[utt54321]

How to log the filename of the file printed by cups?
The filename can be like "/path/to/text.pdf".
I want to print this filename in /var/log/messages.
But the struct cupsd_job_t (schedule/job.h) is very complex for me to find a good point to do so.
Can any ace help me to figure out this stupid question?

[utt54321]

Please someone helps me.
q_q

[zdohnal]

Hi @utt54321 ,

afaik we don't save anywhere the file with absolute path - we only save the filename itself, under name in the cupsd_job_t - it is then used as a job title.

[utt54321]

Hi zdohnal,
1.But I put a txt file (like a.txt) to firefox, and than I use firefox to print it. I found the title is "firefox", where it is not the file's name. Why？
2.If there is nowhere saving absolute path of the printed file, how could cups find it and use it as inputfile? Is there any possibility to find the file's absolute path.

Very thankful for your helping.

[michaelrsweet]

@utt54321 The original filename is not part of the job ticket sent to cupsd. By default, if you "lp /path/to/filename", the default title (job-name) value is the base filename ("filename") but that can be overridden. Many applications will use the document title ("My Great Web Page", "2022 Q2 Finance Report", etc.) and if no title is supplied you'll see "Untitled".

[jsmeix]

Regarding what applications specify as job title
when submitting a print job to CUPS:

This is something that belongs to the individual application.

When the application is a browser:
I experienced a longer time ago that a browser used the URL
as job title which is not a good idea on systems that are used
by more than one user because a URL is personal information
(what sites you visit is personal information)
and the job title is not a secret value so users
may see job titles of other user's print jobs.

[utt54321]

Thank you, sir.
I understand what you say.
If I use a pdf editor to print, how could I know its absolute path in cups?

[jsmeix]

Regarding

path of the printed file, how could cups find it and use it as inputfile

When you print directly on command line via

# lp -d queue_name /path/to/file

the lp program knows /path/to/file
and submits the content of /path/to/file
as print job data via IPP protocol to the cupsd
which stores the print job data in its spool directory
as a new file named /var/spool/cups/d.....-...

So the original file name is not preserved
except e.g. explicitly provided as print job title
when you print via

# lp -d queue_name -t '/path/to/file' /path/to/file

[utt54321]

Very very thankful for your help.

[utt54321]

I have an idea that I can log filename in the cupsDoFileRequest function:

cups/cups/request.c

Lines 41 to 78 in 8456ca2

	ipp_t * /* O - Response data */
	cupsDoFileRequest(http_t *http, /* I - Connection to server or @code CUPS_HTTP_DEFAULT@ */
	ipp_t *request, /* I - IPP request */
	const char *resource, /* I - HTTP resource for POST */
	const char *filename) /* I - File to send or @code NULL@ for none */
	{
	ipp_t *response; /* IPP response data */
	int infile; /* Input file */
	DEBUG_printf(("cupsDoFileRequest(http=%p, request=%p(%s), resource=\"%s\", filename=\"%s\")", (void *)http, (void *)request, request ? ippOpString(request->request.op.operation_id) : "?", resource, filename));
	if (filename)
	{
	if ((infile = open(filename, O_RDONLY | O_BINARY)) < 0)
	{
	/*
	* Can't get file information!
	*/
	_cupsSetError(errno == ENOENT ? IPP_STATUS_ERROR_NOT_FOUND : IPP_STATUS_ERROR_NOT_AUTHORIZED,
	NULL, 0);
	ippDelete(request);
	return (NULL);
	}
	}
	else
	infile = -1;
	response = cupsDoIORequest(http, request, resource, infile, -1);
	if (infile >= 0)
	close(infile);
	return (response);
	}

I changed the DEBUG_printf to be printf:
printf(("cupsDoFileRequest(http=%p, request=%p(%s), resource=\"%s\", filename=\"%s\")", (void *)http, (void *)request, request ? ippOpString(request->request.op.operation_id) : "?", resource, filename));
But I use lp print a file, there is no log in the /var/log showing "cupsDoFileRequest(http=......", why?

[michaelrsweet]

@utt54321 The DEBUG_xxx macros only enable debug logging for the client side. They are not normally compiled in (you need to configure with the "--enable-debug" option) and add a lot of overhead to the libcups APIs. And even if you do build CUPS with that configure option, you need to enable logging with the CUPS_DEBUG_LOG environment variable.

Basically, what you want is not possible.

[utt54321]

Hi michaelrsweet，
I have found a lot of applications supports printing. Firefox and featherpad support printing, but firefox's deb source dose not rely on any cups libs and featherpad seems to use QT api to support printing. They seem to work with cups doing printing job, but I cannot understand that how they send ipp request without using cups libs. How could they work so well?

[zdohnal]

QT implements the communication with cupsd via CUPS API, so it is QT who depends on libcups - featherpad depends on libcups transitively, via QT.
Firefox uses runtime linking - it defines dummy methods with names matching to libcups API and once Firefox is started, it checks for libcups, and if it exists, loads the library and its functions replace the dummy ones.
This way Firefox doesn't have to depend on libcups - several more applications do that with other libs (f.e. Vim doesn't require Python3 libs to narrow its deps...)