> Could somebody please give me a boost?

no need to shout into the void, it is pretty easy to identify the author of manalyzer and ask me directly.

> we are planning to migrate from mana to manalyzer. I started to have a look into it and realized that I have some lose ends.
> Is there a clear migration docu somewhere?

README.md and examples in the manalyzer git repository.

If something is missing, or unclear, please ask.

> Currently I understand it the following way (which might be wrong):
> The class TARunObject is used to write analyzer modules which are registered by TAFactory. I hope this is right?

Please read the README file. It explains what is going on there.

Design of manalyzer had 2 main goals:
a) lifetime of all c++ objects (and ROOT objects) is well defined (to fix a design defect in rootana)
b) event flow and data flow are documentable (problem in mfe.c frontends, etc)

> However, in mana there is an analyzer implemented by the user which binds the modules and has additional routines:
> analyzer_init(), analyzer_exit(), analyzer_loop()
> ana_begin_of_run(), ana_end_of_run(), ana_pause_run(), ana_resume_run()
> which we are using.

I have never used the mana analyzer, I wrote the c++ rootana analyzer very early on (first commit in 2006).

But the basic steps should all be there for you:
- initialization (create histograms, open files) can be done in the module constructor or in BeginRun()
- finalization (fit histograms, close files) should be done in EndRun()
- event processing (obviously) in Analyze()
- pause run, resume run and switch to next subrun file have corresponding methods
- all the "flow" and multithreading stuff you can ignore to first order.

To start the migration, I recommend you take manalyzer_example_root.cxx and start stuffing it with your code.

If you run into any problems, I am happy to help with them. Ask here or contact me directly.

> This part I somehow miss in manalyzer, most probably due to lack of understanding, and missing documentation.

True, I wrote a migration guide for the frontend mfe.c to c++ tmfe, because we do this migration
quite often. But I never wrote a migration guide from mana.c analyzer, because we never did such
migration. Most experiments at TRIUMF are post-2006 and use rootana in it's different incarnations.

P.S. I designed the C++ TMFe frontend after manalyzer and I think it came out quite better, I especially
value the design input from Stefan, Thomas, Pierre, Joseph and Ben.

P.P.S. Be free to ignore all this manalyzer business and write your own analyzer based
on the midasio library:

int main()
{
   TMReaderInteraface* f = TMNewReader(file.mid.gz");
   while (1) {
      TMEvent* e = TMReadEvent(f);
      dwim(e);
      delete e;
   }
   delete f;
}

For online processing I use the TMFe class, it has enough bits to be a frontend and an analyzer,
or you can use the older TMidasOnline from rootana.

Access to ODB is via the mvodb library, which is new in midas, but has been part of rootana
and my frontend toolkit since at least 2011 or earlier, inspired by Peter Green's even
older "myload" ODB access library.

K.O.

> > Is there a clear migration docu somewhere?

I can also give you links to the alpha-g analyzer (very complex) and the DarkLight trigger TDC analyzer (very simple),
there is also analyzer examples of in-between complexity.

K.O.

to continue where we left off in 2019,
https://daq00.triumf.ca/elog-midas/Midas/1520

time to choose the next c++!

snapshot of 2019:

- Linux RHEL/SL/CentOS6 - gcc 4.4.7, no C++11.
- Linux RHEL/SL/CentOS7 - gcc 4.8.5, full C++11, no C++14, no C++17
- Ubuntu 18.04.2 LTS - gcc 7.3.0, full C++11, full C++14, "experimental" C++17.
- MacOS 10.13 - llvm 10.0.0 (clang-1000.11.45.5), full C++11, full C++14, full C++17

the world moved on:

- el6/SL6 is gone
- el7/CentOS-7 is out the door, only two experiments on my plate (EMMA and ALPHA-g)
- el8 was a still born child of RedHat
- el9 - gcc 11.5 with 12, 13, and 14 available.
- el10 - gcc 14.2

- U-18 - gcc  7.5
- U-20 - gcc  9.4 default, 10.5 available
- U-22 - gcc 11.4 default, 12.3 available
- U-24 - gcc 13.3 default, 14.2 available

- MacOS 15.2 - llvm/clang 16

Next we read C++ level support:

(see here for GCC C++ support: https://gcc.gnu.org/projects/cxx-status.html)
(see here for LLVM clang c++ support: https://clang.llvm.org/cxx_status.html)
(see here for GLIBC c++ support: https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html)

gcc:
4.4.7 - no C++11
4.8.5 - full C++11, no C++14, no C++17
7.3.0 - full C++11, full C++14, "experimental" C++17.
7.5.0 - c++17 and older
9.4.0 - c++17 and older
10.5 - no c++26, no c++23, mostly c++20, c++17 and older
11.4 - no c++26, no c++23, full c++20 and older
12.3 - no c++26, mostly c++23, full c++20 and older
13.3 - no c++26, mostly c++23, full c++20 and older
14.2 - limited c++26, mostly c++23, full c++20 and older

clang:
16 - no c++26, mostly c++23, mostly c++20, full c++17 and older

I think our preference is c++23, the number of useful improvements is quite big.

This choice will limit us to:
- el9 or older
- U-22 or older
- current MacOS 15.2 with Xcode 16.

It looks like gcc and llvm support for c++23 is only partial, so obviously we will use a subset of c++23 
supported by both.

Next step is to try to build midas with c++23 on el9 and U-22,24 and see what happens.

K.O.

> After starting midas (mhttpd &, and mlogger -D) and running the `fetest` frontend I went into the midas/python/examples directory and ran basic_hist_script.py, and, even though I could see the 'pytest' program in the Programs page,
> 
> Valid events are:
> Enter event name:
> 
> was printed out, which signified that no events were found. No errors were displayed.

To check that MIDAS itself is built correctly, you can try "make test", this will create a sample experiment,
run fetest, start, stop a run and check that data file and history file is created with correct history events.

If "make test" fails, I can help debug it.

In your experiment, you can check that history files are created correctly:

1) "mhist -l" should show all available events
2) "mhdump -L *.hst" should show all events in the .hst history files
3) if you have the newer mhf*.dat files, you can "more mhf_1449770978_20151210_hv.dat" to see what data is inside

If all of that works as expected, there must be a problem with the python side and we will have to figure
out how to reproduce it.

This reminds me, "make test" does not test any of the python code, it should be added (and python should be added
to the bitbucket builds).

K.O.

>
> However right after running these commands I removed a .SHM_HOST.TXT file
>

Instead of deleting .SHM_HOSTS.TXT, please create it as an empty file. I thought the documentation is clear about it?

Also we recommend installing MIDAS to $HOME/packages/midas. There is a number of problems if installed at top level.

If you want to be compliant with the Linux LFS, /opt/midas is also a good place.

> 
> ... and it suddenly worked!
>

We still did not establish if mhist, mhdump and the other commands I sent you work correctly,
to confirm MIDAS is creating correct history files. (before you try to read them with python).

Also we did not establish that you have correct paths setup in ODB /Logger/History.

Many things can go wrong.

Next time python history malfunctions, please do all those other things and report to us. Thanks!

K.O.

We have a problem with over aggressive AI bots. They are scanning everything and 
are putting a crazy load on the mediawiki mysql database. This makes all out 
wikis very slow and unresponsive, which is a big problem.

These AI scanners do not seem to know what they are doing and are trying to load 
all the special pages, like "what links here" and "recent changes" and complete 
revision histories for each page. I am not impressed. Old school google and bing 
scanner bots are much more respectful and do not cause problems.

AI bots are also scanning this elog, and to Stefan's credit elogd is holding the 
load just fine.

To reduce load on the MidasWiki mysql database, I now restricted access to most 
special pages to logged in users. It seems to have helped.

If this causes big difficulty or if you have suggestion on better ways to deal 
with aggressive AI scanner bots, please speak up.

One alternative is to put MidasWiki behind a generic login page with a well 
known password. Downside is it will block google and bing search engines, too.

K.O.

> Does anyone have experience with writing a MIDAS frontends to communicate with a device that operates using LabView (e.g. superconducting magnets, cryostats etc.). Any information or experience regarding this would be highly appreciated.

Yes, in the ALPHA anti-hydrogen experiment at CERN we have been doing this since 2006.

Original system is very simple, labview side opens a TCP socket to the MIDAS felabview frontend
and sends the numeric data as an ASCII string. The first four chars of the data is the name
of the MIDAS data bank, second number is the data timestamp in seconds.

LCRY 1234567 1.1 2.2 3.3

A newer iteration is feGEM written by Joseph McKenna (member of this forum), it uses a more sophisticated
labview component. Please contact him directly for more information.

I can provide you with the source code for my original felabiew (pretty much unchanged from circa 2006).

K.O.

> beta version of macOS: Switching beta updates off fixed the issue.

I would be very surprised if that was the problem.

Bigger concern is that it fails without producing any useful error message.

Latest MacOS makes it extremely difficult to debug this kind of stuff, there
is several hoops to jump through to enable core dumps and to allow lldb
to attach and debug running programs.

K.O.

> Thanks Konstantin. Please send me the felabview code or let me know where I can find it.

https://bitbucket.org/expalpha/a2daq/src/alpha/src/felabview.cxx

this is code circa 2006, there are now better ways to do some of that coding.

if you want bidirectional communication with labview, read/write odb, etc, simplest is probably
to write the "mjserver" that talks midas json-rpc over plain tcp, without all the http/https
gunk you need to go through mhttpd.

K.O.

we are migrating the dragon experiment from an old mac to a new mac studio and we ran into a problem 
where one equipment format was set to "fixed" instead of "midas". lots of confusion, mdump crash, 
analyzer crash, etc. (mdump fixes for this are already committed).

it made us think whether equipment format is still needed. in the old days we had choice of MIDAS and 
YBOS formats, but YBOS was removed years ago, and I was surprised that format FIXED was permitted at 
all.

I did a midas source code review, this is what I found:

- remnants of YBOS support in a few places, commit to remove them pending.
- FORMAT_ROOT is used in mlogger for automatic conversion of MIDAS banks to ROOT trees
- FORMAT_FIXED is used in a few slow control drivers in drivers/class, instead of creating MIDAS 
banks, they copy raw data directly into an event (there is no bank header and no way to identify such 
events automatically)
- lots of code to support different formats in mdump (mostly dead code)
- the rest of the code does not care or use this format stuff

Current proposal is to remove support for all formats except FORMAT_MIDAS (and FORMAT_ROOT in 
mlogger).

- defines of FORMAT_XXX will be removed from midas.h
- "Format" will be removed from ODB Equipment/Common
- "Format" will be removed from ODB Logger/Channel
- to maintain binary compatibility, we can keep the "Format" ODB entries, but they will be ignored.

List of slow control drivers that support FORMAT_FIXED:

daq00:midas$ grep FORMAT_FIXED drivers/class/*
drivers/class/cd_fdg.cxx:   if (fgd_info->format == FORMAT_FIXED) {
drivers/class/cd_ivc32.cxx:   if (hv_info->format == XFORMAT_FIXED) {
drivers/class/cd_rf.cxx:	if (rf_info->format == XFORMAT_FIXED) 
drivers/class/generic.cxx:   if (gen_info->format == XFORMAT_FIXED) {
drivers/class/hv.cxx:   if (hv_info->format == XFORMAT_FIXED) {
drivers/class/multi.cxx:   if (m_info->format == XFORMAT_FIXED) {
drivers/class/slowdev.cxx:   if (gen_info->format == XFORMAT_FIXED) {
daq00:midas$ 

K.O.

> ALL MIDAS CLIENTS GET RE-COMPILED after the new code is applied.

Usually we expect that it is "safe" to update to the latest version of MIDAS.

In the sense that we do not have to track down every single frontend
and rebuild it. We have several experiments where tracking down the source code
and rebuilding a frontend takes more than 5 seconds. In many cases these are
10 year old executables that worked just fine through many updates of MIDAS
without having to rebuild them.

So any binary incompatible change is best avoided and must be clearly announced.

The present binary-incompatible change is this commit:
https://bitbucket.org/tmidas/midas/commits/5899f1657ba31121c9f420a824b3c6c13b173988

I tagged the last commit before this change as: midas-2024-12-a

K.O.

Unnamed person who added this clever bit of c++ coding, please fix this compiler warning. Stock g++ on Ubuntu LTS 24.04. Thanks in advance!

/home/olchansk/git/midas/src/system.cxx: In function ‘std::string ss_execs(const char*)’:
/home/olchansk/git/midas/src/system.cxx:2256:43: warning: ignoring attributes on template argument ‘int (*)(FILE*)’ [-Wignored-attributes]
 2256 |    std::unique_ptr<FILE, decltype(&pclose)> pipe(popen(cmd, "r"), pclose);

K.O.

I am getting a scary compiler warning from MSCB code. I generally avoid touching MSCB code because I have no MSCB hardware to test it, so I am 
asking the persons unnamed who wrote this code to fix it. Stock g++ on Ubuntu LTS 24.04. Thanks in advance!

In function ‘ssize_t read(int, void*, size_t)’,
    inlined from ‘int mscb_interprete_file(const char*, unsigned char**, unsigned int*, unsigned char**, unsigned int*, unsigned char*)’ at 
/home/olchansk/git/midas/mscb/src/mscb.cxx:2666:13:
/usr/include/x86_64-linux-gnu/bits/unistd.h:28:10: warning: ‘ssize_t __read_alias(int, void*, size_t)’ specified size 18446744073709551614 
exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
   28 |   return __glibc_fortify (read, __nbytes, sizeof (char),
      |          ^~~~~~~~~~~~~~~
/usr/include/x86_64-linux-gnu/bits/unistd-decl.h: In function ‘int mscb_interprete_file(const char*, unsigned char**, unsigned int*, unsigned 
char**, unsigned int*, unsigned char*)’:
/usr/include/x86_64-linux-gnu/bits/unistd-decl.h:29:16: note: in a call to function ‘ssize_t __read_alias(int, void*, size_t)’ declared with 
attribute ‘access (write_only, 2, 3)’
   29 | extern ssize_t __REDIRECT_FORTIFY (__read_alias, (int __fd, void *__buf,
      |                ^~~~~~~~~~~~~~~~~~

K.O.

bitbucket automatic builds were broken after mfe.cxx started printing some additional messages added in commit
https://bitbucket.org/tmidas/midas/commits/0ae08cd3b96ebd8e4f57bfe00dd45527d82d7a38

this is now fixed. to check if your changes will break automatic builds, before final push, please do:

make clean
make mini -j
make cmake -j
make test

K.O.

Some time ago Ben Smith added test coverage reports using GCC -fprofile-arcs -
ftest-coverage and lcov.

It reports code coverage after running "make test". Reported code coverage is 
surprisingly high for the very little code executed by "make test" - create ODB, 
start a frontend,  start run, stop run, check that mlogger created data files 
and history files.

(Of course coverage can never reach 100%, some obscure branches are unreachable 
without using an automated fuzzer, and some error paths are unreachable without 
also using a system call fault injector).

Simplest way to generate a coverage report:

make clean
make cmake YES_COVERAGE=1
make coverage
open cov_html/index.html

K.O.

> currently to link Midas to project one has to do several steps ...

this information is incorrect. please read https://daq00.triumf.ca/elog-midas/Midas/2258

a very simple way to use link MIDAS using midas-targets.cmake has been implemented a long time ago.

before proposing a new way of doing things, it would be nice to hear about shortcomings
of the existing stuff. A simple "Konstantin's way sucks" or "this is not the cmake way!"
would have been sufficient.

K.O.

> After some iterations, we merged the branch with the new build scheme.

the commit to implement this change in the manalyzer was not pushed, for reasons unknown.

fixed, commit f2b4dc87ca4830f6bed8667d6a4ee4afd6d242a1

K.O.

> I have a manalyzer that uses a derived class of TMFeRpcHandlerInterface to communicate information to 
> Midas during online running.  At the end of each run it saves out custom data in the 
> TMFeRpcHandlerInterface::HandleEndRun override. This works really well.
> However, when I run offline on a Midas output file the HandleEndRun method is never called and my data is 
> never saved.  Is this intentional?  I understand that there is no point for the HandleBinaryRpc method offline, 
> but the other methods (HandleEndRun, HandleBeginRun etc) could serve a purpose.  Or is it a conscious 
> choice to ignore all of TMFeRpcHandlerInterface when offline?

apologies for delayed response.

I saw the question, completely did not understand it, only now got around to figure out what is going on.

according to manalyzer/README.md, section "manalyzer module and object life time", BeginRun() and EndRun() is called 
always. Offline and online. What you see would be a bug that we do not see in our environment. I confirm this by 
running manalyzer in a demo mode: ./bin/manalyzer_test.exe  --demo -e10 -t

no, wait, you say you use HandleBeginRun() and HandleEndRun(). this is not right, they are not part of the manalyzer 
API and they indeed are only used when running online.

correct solution would be to use BeginRun() and EndRun() instead of HandleBeginRun() and HandleEndRun().

you could also save your data in the module destructor (although good programming recommendation is to use 
destructor only for unavoidable things, like freeing memory, etc).

K.O.

I think I added the cache size correctly:

  {"Trigger",               /* equipment name */
      {1, 0,                 /* event ID, trigger mask */
         "SYSTEM",           /* event buffer */
         EQ_POLLED,          /* equipment type */
         0,                  /* event source */
         "MIDAS",            /* format */
         TRUE,               /* enabled */
         RO_RUNNING |        /* read only when running */
         RO_ODB,             /* and update ODB */
         100,                /* poll for 100ms */
         0,                  /* stop run after this event limit */
         0,                  /* number of sub events */
         0,                  /* don't log history */
         "", "", "", "", "", // frontend_host, name, file_name, status, status_color
         0, // hidden
         0  // write_cache_size <<--------------------- set this to zero -----------
      },
   }

K.O.

the main purpose of the event buffer write cache is to prevent high contention for the 
event buffer shared memory semaphore in the pathological case of very high rate of very 
small events.

there is a computation for this, I have posted it here several times, please search for 
it.

in the nutshell, you want the semaphore locking rate to be around 10/sec, 100/sec 
maximum. coupled with smallest event size and maximum practical rate (1 MHz), this 
yields the cache size.

for slow control events generated at 1 Hz, the write cache is not needed, 
write_cache_size value 0 is the correct setting.

for "typical" physics events generated at 1 kHz, write cache size should be set to fit 
10 events (100 Hz semaphore locking rate) to 100 events (10 Hz semaphore locking rate).

unfortunately, one cannot have two cache sizes for an event buffer, so typical frontends 
that generate physics data at 1 kHz and scalers and counters at 1 Hz must have a non-
zero write cache size (or semaphore locking rate will be too high).

the other consideration, we do not want data to sit in the cache "too long", so the 
cache is flushed every 1 second or so.

all this cache stuff could be completely removed, deleted. result would be MIDAS that 
works ok for small data sizes and rates, but completely falls down at 10 Gige speeds and 
rates.

P.S. why is high semaphore locking rate bad? it turns out that UNIX and Linux semaphores 
are not "fair", they do not give equal share to all users, and (for example) an event 
buffer writer can "capture" the semaphore so the buffer reader (mlogger) will never get 
it, a pathologic situation (to help with this, there is also a "read cache"). Read this 
discussion: https://stackoverflow.com/questions/17825508/fairness-setting-in-semaphore-
class

K.O.