When an unauthenticated user accesses a protected resource of my API server, I want to immediately redirect them to the login page.  Research has indicated that the server should send the client an HTTP 302 page with the url.  Experimentation has shown that you can successfully set cookies. The cookie can be used to hold the url to re-redirect the user back to the original request after authenticating. Nice.

Here’s the header to send from the server:


    const string cstr_HTML_302_HEADER1 = "HTTP/1.1 302 Moved Temporarily\r\nLocation: ";
    const string cstr_HTML_HEADER2 = "\r\n\r\n";
    // ....
        string cookie_header = "\r\n";
        cookie_header += "Set-Cookie: .... ";
        *response << cstr_HTML_302_HEADER1 << "/v1/login.html" << cookie_header << cstr_HTML_HEADER2;

The publicly available version of quick-http on github needs a refresh, hopefully I'll have some time soon to move a large set of new code there that uses Simple Web Server, HTTP 302 redirects, user authentication, etc.

A friend of mine asked for help to do a complicated data pivot on a few thousand rows of data.  We discussed it and agreed that we probably needed to walk each row and do an algorithmic summation on the fly.

My first instinct was to fire up Windows and dust off the old VBA cobwebs in my mind.  Fortunately that didn’t last long, and I reached for a far more comfortable friend, SQL.  The final SQL wasn’t terribly simple, but wasn’t terribly complicated either.  And it was a total linux solution, always the more pleasant solution.  And totally gui, too, for all you folks who get hives when you have to bang at a terminal.  🙂

  • Open excel sheet in LibreOffice
  • Copy columns of interest into a new sheet
  • Save As -> Text CSV
  • Create a sqlitebrowser database
  • Import the CSV file as a new table

BAM, easy, powerful, free, cheap, fast, win, win, win!

I ended up using this sql, the key is sqlite’s group_concat(), which just jams a whole query into one field, ha, boom.

SELECT a.email
 , group_concat(d.Dates) AS Dates
 , group_concat(d.Values) ASValues
FROM (select * from TestData ORDER BY email,Dates) AS a
inner JOIN TestData AS d ON d.email = a.email
GROUP BY a.email;

Then I copied the tabular result of the query, and pasted into the LibreOffice sheet, which asked me how to import.  Delimited by tab and comma, and then always recheck “select text delimiter” and use a comma.  And the cells fill right up.

All is full of light.

Simple-Web-Server has turned out to be a perfect fit for A better Trader… more to come…

Ubuntu has a few cross compile toolchains available, for example:

sudo apt install gcc-arm-linux-gnueabi # for arm chips
sudo apt install gcc-arm-linux-gnueabihf # for arm chips with floating point registers

Crosstool-NG is a great help in getting the tougher cross compile toolchains built. I was targeting a specific ARM 926EJ-S chipset, and the vendor had their own toolchain.

Trying to cross compile on newer versions of ubuntu was a headache though. I would get “FATAL: kernel tool old” errors from the executable on the target system, and the autotools configure option to limit kernel version (–enable-kernel=…) did not fix it.

In the end, I was able to build zlib and openssl and libwebsockets statically, with a Crosstool-NG toolchain on an older ubuntu, but had to deal with this glibc bug by explicitly specifying the system lib folder, or getaddrinfo, gethostbyname, and all other NSS based functions just plain old failed (lost some hair to that fine feature – cmon RMS and Drepper, get along better so your code doesn’t stink…):

LD_LIBRARY_PATH=/lib ./my_static_app

Details as usual on the wiki. Onwards.

I love this project introduction because it perfectly describes the essence of agile development without using a single word of marketing horse manure.