For some games the API shows files with "count" and "total_size" of zero meaning that they are not available to download
These are listed on account page for information purpose only (for example to show DLCs included in the installer)
Feature request by "coffeecup" on GOG forums
Useful for sorting big libraries with --subdir-game option
Takes the first letter from %gamename%
If %gamename% begins with a number then 0 (zero) is used instead
Makes downloader use duplicate handler for extras
This should help to avoid a race condition mentioned in issue #200 (Crashes when downloading "secure") because the downloader no longer tries to download the same file from multiple threads
Sometimes the download fails with CURLE_SSL_CONNECT_ERROR and retrying
is successful.
Sometimes multiple retries are needed, but that is already handled by
the existing code.
The variable name "current_changelog" was ambiguous. It could apply to both the newly obtained changelog from GOG and the previously saved changelog on the filesystem.
I accidentally reversed the return code for Downloader::downloadFileWithId when rewriting it to use Galaxy API in a884e8c0a3ebb66af43f514bb03e23e67aecead6
This changes the success return code back to 0 like it was previously.
Fixes issue with filenames containing ampersand.
Also fixes issue with --download-file not assigning proper filename automatically.
New method detects the path more accurately when encountering the currently used url format but in case of unknown url format it fallbacks to the old method.
Previously orphan checking was limited to user selected language and platform values.
This meant that by default only English language and Windows+Linux was selected.
Overriding the global config values and clearing previously got details before starting the orphan check makes --check-orphans work as intended.
There was no reason to use the old downloader API for this because we can get the same data using Galaxy API.
This also means that Downloader::downloadFileWithId is now the only function using the old API.
This changes the behavior so that we save XML data before skipping complete file.
Previously we skipped to next file right after determining that the file we were attempting to download was already complete.
This caused issues when user changed XML directory with --xml-directory option after already downloading some files.
The new XML directory would never have XML data for complete files but because we skipped before saving XML data.
Adds option --enable-login-gui which is used to enable GUI login when reCAPTCHA is encountered on login form.
This reverses the previous behavior added in ea0ec2a9bdac3ac2e7446dbfb2189a6a141e512a and removes --disable-login-gui option.
It's better to have GUI login disabled by default just in case user is running lgogdownloader through SSH without X forwarding.
Because user has to enable login GUI first there won't be any unexpected behavior.
Make login GUI automatically fill email and password fields
Users no longer need to type email and password again just because normal login method failed due to reCAPTCHA and GUI was launched
Adds option --disable-login-gui which is used to disable GUI login when reCAPTCHA is encountered on login form.
Useful when downloader is compiled with GUI support but user is using it through SSH without X forwarding.
This patch fixes issues with some calls to the rhash C library that
cause subtle memory corruptions. There are two main issues that this
patch addresses:
1) The library init function is called multiple times in various places
within in the Util class. This is not necessary and may cause issues
if the initialization is repeated simultaneously in multiple threads.
Therefore the call was moved to the beginning of the main() function.
2) There is not enough space in the output buffers to store the
terminating NUL character of the hex-formatted hashes. The print
function from the rhash library writes a regular C string to
the output buffer and expects enough space to include the end
marker.
Depending on the memory layout generated by the compiler, this results
in one byte of the stack to be overwritten, which might lead to random
issues that are hard to find. On AArch64 (ARM) with GCC 6.3 the call to
the rhash_print() function happens to destroy the lower byte of the
frame pointer and causes a crash due to an invalid free() operation some
time after the Util::createXML function returned.
