changes

seed格式数据解压程序,地震分析人员必备

13 apr 90	bugfix	mw

the command: output xxx yyy 4096 32768 0,30 0,1

failed with a msg, create_T: nlvol not zero. This meant
a mismatch between the timespan count and the number int
the linked list. The cause turned out to be a bad address
in output_tspan, in the code which checked how many T type
logical records to output. It output them all in the first tspan.

Also, there was an internal assumption that the end timespan
would be start tspan + ntspan*tspan_break. This is not true
if there are holes in the request (as there were). See the
comments in create_T.c and create_V.c about this. They now
search for a  start/end tspan which brackets the start time
of the next data to be accepted (in the lvol linked list).

** This may not be the way that IRIS will always want to do it **

Probably the tspan_list structure should be amended to contain
the start and end times within each node - it would make all of
this much more simple, but it needs a bit more checking to implement.

struct tspan_list
{
	int					count;
	int					seqno;
	struct	input_time	start;	/* start time for this tspan */
	struct	input_time	end;	/* this too just in case tspans are not all
								 * the same size ?? */
	struct	lvol		*start;
	struct	tspan_list	*next;
};

usage.c was updated to note that output disk files are *appended*
to, which might be unexpected behavior. Rdseed will break if
the same output file is used twice.


A bug in the way that the volume logical record, end_volume time
was computed in create_V.c was fixed. It was setting the end_volume
time to the start_volume + ntspan*tspan_break if tspan_break was
non-zero. If the volume break is zero, it now prints an error msg
and dies, since the sum of tspans volume span is wrong if holy
tspans are used.


15may90 mw	enhancement

SEED v2.1 has several new blockettes which reduce the amount of
information redundantly included. These are 043-048 abbreviation
blockettes, and 060, which references them. Include code in
Header_phase/{ load_data.c, is_in_twindow.c } which permits this.
Also clean up load_data.c (getmore), which was calling big_buffer
more often than necessary.

PROBLEM: these blockettes should not appear on a V2.0 SEED volume.
We can't yet call output a v2.1 writer, since it does not write 074
blockettes yet.

12jul90 mw	bugfix

Channels with sample interval = 0 were being ignored even though
they were requested and seemed to be in the window.

The problem was that the end time of a data record was computed
as the start time, due to delta-t = 0. This is fixed by passing the
end timespan time down from the main into the routine which computes
this end time. If the sample int is zero, the tspan time is used rather
than start+deltat*nsamples.

Main/main.c	Scan/{find_data.c,find_segment.c,get_scaninfo.c}


16jul90 mw		BUGFIX	more timespan screwups

Sigh. A typo in two strncmp() calls which determined when the
station/locations/channel changed while scanning the lvol list
caused a fatal core dump. The incorrect character count was
5+7+3, and it should have been either 5+2+3 *or* 7+3.
It would have additionally compared the year and one byte of the
day in the data record header.

fix problem with large (n=255) datapeices. This generates blockettes
about 13k in size (if it could). Breaking blockettes into 9999 byte
pieces broke during the fix of 04/13/90 (was that a Friday?)

fix problem with last pointer in each 073 - was not adjusted for
the size of the logical records comprising the timespan headers.

Header_phase/create_T.c was modified.


09 Nov 90 mw

change debugging printout level of all subroutines in Utilities
from >= D_MAX to > D_MAX - I'm tired of paging through
all the crap. Will do same to other routines later...

BUGFIX:

The problem seems to be that when triggered data is not in the
user window, there may be no data in the timespan, even though
many user reqeust records were processed. For example, in the
end of req.temp,

[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 1
end_tspan       1989    071     00:00:00.0000
[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 1
[find_data] Nodes added to lvol = 0
end_tspan       1989    072     00:00:00.0000
end_tspan       1989    073     00:00:00.0000
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 0
[find_data] Nodes added to lvol = 0

day 72 has no requests for it, but 73 has requests, but there
is no data which is within the window. This blows up create_T
which tries to make a tspan for a non-existent time span!

This was fixed by having both create_T.c and create_V.c check
for a count of 0 datachunks in a timespan. This will only occur
if the last timespan in a volume found no data records in a window.
(It is ok if there is an empty tspan in the middle of a volume -
eventually it will either have something in it, or it will be the
last tspan in a volume and so be caught by the above test.)
Flush_output_volume.c also now checks for this trailing empty
timespan.

Along the way, a bug which caused create_T.c to mistakenly
omit blockette 070's in a timespan was fixed. The problem
here was the improbable assumption that having the same
station/channel/location together in the lvol list meant
that these were the same "datapeices" for 073 blockette
purposes.

The incorrect version was gluing together datapieces even
across timespan boundaries. This is fixed by downcounting
the count of datachunks (lvol nodes) within each timespan.
It only worked before because there were always a mix of
station/channels in adjacent timespans. Sigh. The data records
were *not* affected, only the timespan headers.

create_T.c was neatened up some by adding new routine make_room.


19-nov-90 mw	begin changes for SEED v2.1

changed SEED_VERSION to SEED_DEFAULT ("02.0")
added new routine get_SEED_version which sets SEED_Version to
	environment variable SEED_VERSION's float value
coding in create_v.c and create_T.c will be done to detect
	value of SEED_Version and respond accordingly.

20-nov-90 mw	continue seed v2.1 changes

	create_V.c	use SEED_Version instead of SEED_DEFAULT (see above)
	create_T.c	add if-then-elses based on SEED_Version and blockette_type
				completely untested yet.

02dec90	mw		continue seed v2.1 changes

	create_V.c				leading zero in format ("02.1", not " 2.1")
	create_T.c				get the 074 pointers right, get tspan numbering
							right
	flush_output_volume.c	put all tspans at front


24dec90 mw		more robustness

	get_scaninfo.c			if sample rate factor is non-zero, but sample
							rate multiplier is zero, assume that the multiplier
							should be one. This is undefined by the seed
							document. Should it be an error?



19mar91	mw		problems with "non-existent file"s which really exist

	find_data.c				in the case that a request record finds
							no data in the file, in one case, this routine
							would return without closing the open file.
							Eventually, open() would get a "too many
							files open" error. This was fixed, and better
							error messages for debugging added.

							This particular problem occurred because the
							end time was earlier than the start time in
							several bad Holdings.File records.


03apr91 mw		better appearance

	Main/main.c				remove "untested v2.1" message

	Main/usage.c			better description of SEED time fields

16apr91 mw		bugfixes

	Header/
		get_chnl_hdr.c		zero last byte of atoi() strings clen,ctype
		load_data.c			zero last byte of atoi() strings clen,ctype
		make_lrec.c			zero last byte of atoi() strings clen,ctype
	Utility/
		big_buffer.c		change scheme for allocation to prevent extra
							block of allocation.

	tested on 4/330 under 4.1 in combinations of <none>, -p, -g, -O.
	tested on 3/50 under 4.1; 3/280 under 3.5; 4.280 under 4.1.


06 june 91	mw			change


	Includes/structures.h
			increase the size of the file name remembered in the lvol
			node from 40 chars to 256 chars so nobody can ever again
			build a directory/filename which is too long. So long as
			we always have lots of virtual space, 10,000 name requests
			will only eat 2.5 Mbyte of virtual. Traversing this list
			is usually done in io bound loops, so the increased paging
			should not be a factor. Also move the name to the end of
			the structure, to reduce potential padding problems.

17 june 91	mw			change

	Scan_phase/find_data.c

			check the offsets found by find_segment. If they are
			impossible ( <0, end>=start, >size of file ) then barf
			with a FATAL bugcheck. Prompted by such case (how?) in
			Scripps testing data.

18 june 91	mw			bugfix (sort of)

	Scan_phase/find_data.c
		Warn if the file's actual size is not consistent with that
		given by the request record, but use the actual.

		Always set the expected number of segments to zero, as the
		code within find_segment should never quit checking for
		segments.

	Scan_phase/find_segment.c

		More debugging included; this routine needs a rewrite badly
		and should be simplified.


03 dec 91 mw		bugfixes from Sue Schoch, made in ?june 91

	Scan_phase/find_data.c

		don't warn about file offsets which seem to be incorrect;
		find_segment can in some cases return none. find_segment
		should probably be fixed to avoid this.

	Scan_phase/find_segment.c

		increase the size of msg[] for long file names and seed times.

04 dec 91 mw	changes for v2.2 SEED started

	Main/main.c					add call to get_creator()
	Main/usage.c				document environment variable CREATOR
	Utilities/get_creator.c		new routine, needed for new blockette 10 field
	Header_phase/create_V.c		change code for blockette 10
	Header_phase/make_headers.c	change assumption that blockette 10 size fixed

	change all references from floating point SEED_Version to
	integer *10 SEED_Version; thus, version 2.2 appears as 22.
    	Header_phase/create_T.c, create_V.c, make_headers.c
		Include/output.h, globals.h
		Main/main.c
		Output_phase/flush_output_volume.c

07 dec 91 mw	more changes for v2.2 SEED

	Include/structures.h	new fields for lvol: gofast and ngofast
		for eventual use by Scan_phase stuff to find accelerator indicies
	Scan_phase/put_lvol.c	zeroes out both new fields.

09 jan 92 mw	bugfix

	Scan_phase/get_scaninfo.c

	symptoms: when a slow channel (ANMO UHE) data record crossed a
	year boundary (90 - 91) and the start time was late in the day
	(12:00) the calculation of the end time in get_scaninfo did not
	take into account the possibility of overflow in add_time. Up to
	one day of time might have to be added in, this could cause
	overflow.

10 jan 92 mw	more problems

	also a problem: when ultra low sample int channels have 2 records,
	say 2 4k records, then the start time of the second chunk is
	about 3 days after the start of the 1st chunk. If you use 1 day
	tspans, then the second chunk is 3 tspans after higher rate channels
	which surround it. This causes "holy" tspans


11 feb 92 mw	bugfixes

	Scan_phase/get_segment.c

	for SEED v2.1 or greater, blockette 074's had end times which came
	from the next record's end times, typically. This has been fixed.
	No other symptoms were observed for this bug.

14 sep 92 wan	cosmetic

	Header_phase/get_stn_path.c

	Print "No station comments available" message only when above
	D_MIN debugging level.
	
15 sep 92 wan   bugfix

	Header_phase/build_stn_path.c

	This function did not allow station name "PFO_T" to be processed.
	Changed station name QC to allow the "_" character in station names.

sept, 93 chris L
	
	added support for block 71s being in the data_files or header 
	directories.
	Fixed problem with outputing DATALESS seed
	added support for seed version 2.3 including SEED_LABEL
    fixed bug in make_room
	added check for byte swapped data. Timing data in data records are
	integers.
	changed program name from "output" to "pod"


Aug 31st, 94 - added checking for garbage in the data files.

Mar 1st 1995 - changed the volume beg end time to really be the
		  	   beginning and ending time span
	
Mar 15th 95  - updated volume start end time for dataless seed. Since no
				timespans available, use the request start,end

MAr 15th 95  - made pod "network" aware, no longer dependent upon unique
				station names within volume. New HAR... station/channel
				naming convention - took out 'S' & '###' & 'C'

May of 97   - support for varying record lengths, from passcal data

Dec-Feb -97,
Spring 98      - support for location code processing
	       - added PODJUSTDOIT functionality
	       - some bug fixes, bug when user entered logical record 
		 size less then 4k 
	       - bug with processing LOG records
	       - bug fix for multi log record processing. pod sometimes was
		    off slightly when calculating B74s.

POD 4.0.b1
Sept 30 1998   - changed the h file, took out old and unused fields.
	       - final version of location code processing
		    - new channel directory naming convention, ala LHE.03

POD 4.0.b2
Nov 2, 1998    -took out BW from holdings file
	

POD 4.1 	added in support for Blockette 42 and 62
		added ability to scan data directories recursively.
			No need to specify each and every data
			directory in the environment

		added in PODJUSTDOIT functionality to DATALESSes

		small bug, check for sample rate of zero in 
			get_scan_info().

POD4.1.2 	small bug in data record length scan
		added in rec->nsamples == 0 test (from UCB)

podv4.1.5  	
Sept 27 2002	small bug in scan_info.c. When calculating the
		duration of a record, it used nsamples-1, which 
		caused problems. Took out the -1.
podv4.2
March 3 2004	Redid how pod outputs sequential record numbers.
		Sequential record numbers now exactly mirror the
		logical record.

			data records < log rec size now get built
			with sequential record numbers == log rec count
		
		Changes some warnings to ouput the stn/chn info on
		same line as the warning label